derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
29693 lines
1.0 MiB
Plaintext
29693 lines
1.0 MiB
Plaintext
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX900 %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX90A %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX942 %s
|
|
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX9-LABEL: v_shuffle_v2i64_v8i64__u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> poison
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX9-LABEL: v_shuffle_v2i64_v8i64__8_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_u(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 poison>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__15_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__15_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__15_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__15_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> zeroinitializer
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_0(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 0>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_1(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 1>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_2(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 2>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_3(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v23, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[20:23], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v23, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[20:23], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 3>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v23, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[20:23], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v23, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[20:23], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_4(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v25, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[22:25], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v25, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[22:25], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 4>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v23, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[20:23], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v23, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[20:23], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v25, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[22:25], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v25, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[22:25], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_5(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v27, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[24:27], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v27, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[24:27], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 5>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v19, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[16:19], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v19, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[16:19], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v23, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[20:23], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v23, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[20:23], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v25, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[22:25], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v25, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[22:25], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v27, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[24:27], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v27, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[24:27], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_6(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v29, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[26:29], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v29, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[26:29], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 6>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v21, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[18:21], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v21, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[18:21], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v23, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[20:23], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v23, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[20:23], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v25, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[22:25], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v25, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[22:25], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v27, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[24:27], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v27, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[24:27], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v29, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[26:29], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v29, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[26:29], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_7(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v31, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[28:31], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v31, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[28:31], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 7>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX9-LABEL: v_shuffle_v2i64_v8i64__u_8:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX9-LABEL: v_shuffle_v2i64_v8i64__8_8:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_8(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 8>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v3
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_9(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 9>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_10(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 10>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v20
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v21
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v21
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v21
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v22
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v22
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_11(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 11>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v20
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v21
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v21
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v21
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v22
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v23
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v22
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v23
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v22
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v23
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v24
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v25
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v24
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_12(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 12>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v20
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v21
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v21
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v21
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v22
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v23
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v22
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v23
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v22
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v23
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v24
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v25
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v24
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v25
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v24
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v25
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v26
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v26
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_13(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 13>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v20
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v21
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v21
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v21
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v22
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v23
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v22
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v23
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v22
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v23
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v24
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v25
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v24
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v25
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v24
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v25
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v26
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v27
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v26
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v27
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v26
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v27
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v28
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v29
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v28
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v29
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_14(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 14>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__u_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__u_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__u_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__u_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__0_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__0_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:17]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v16
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v17
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v4, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__0_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:17]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v17
|
|
; GFX90A-NEXT: global_store_dwordx4 v18, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__0_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v18, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:17]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v16
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v17
|
|
; GFX942-NEXT: global_store_dwordx4 v18, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__1_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__1_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v18
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v19
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__1_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v18
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v19
|
|
; GFX90A-NEXT: global_store_dwordx4 v20, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__1_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v20, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v18
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v19
|
|
; GFX942-NEXT: global_store_dwordx4 v20, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__2_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__2_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:21]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v20
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v21
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__2_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:21]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v20
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v21
|
|
; GFX90A-NEXT: global_store_dwordx4 v22, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__2_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v22, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:21]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v20
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v21
|
|
; GFX942-NEXT: global_store_dwordx4 v22, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__3_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__3_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v22
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v23
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__3_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v22
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v23
|
|
; GFX90A-NEXT: global_store_dwordx4 v24, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__3_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v24, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v22
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v23
|
|
; GFX942-NEXT: global_store_dwordx4 v24, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__4_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__4_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[10:25]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v24
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v25
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__4_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[10:25]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v24
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v25
|
|
; GFX90A-NEXT: global_store_dwordx4 v26, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__4_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v26, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[10:25]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v24
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v25
|
|
; GFX942-NEXT: global_store_dwordx4 v26, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__5_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__5_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v26
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v27
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__5_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v26
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v27
|
|
; GFX90A-NEXT: global_store_dwordx4 v28, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__5_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v28, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v26
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v27
|
|
; GFX942-NEXT: global_store_dwordx4 v28, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__6_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__6_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[14:29]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v28
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v29
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__6_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[14:29]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v28
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v29
|
|
; GFX90A-NEXT: global_store_dwordx4 v30, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__6_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v30, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[14:29]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v28
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v29
|
|
; GFX942-NEXT: global_store_dwordx4 v30, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__7_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__7_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v17, v15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[14:17], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__7_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, v30
|
|
; GFX90A-NEXT: v_mov_b32_e32 v17, v31
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v32, v[14:17], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__7_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v32, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, v30
|
|
; GFX942-NEXT: v_mov_b32_e32 v17, v31
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v32, v[14:17], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__8_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__8_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__8_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__8_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__9_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__9_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__9_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__9_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__10_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__10_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__10_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__10_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__11_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__11_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__11_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__11_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__12_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__12_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__12_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__12_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__13_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__13_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__13_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__13_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v15
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v2i64_v8i64__14_15(ptr addrspace(1) inreg %ptr) #0 {
|
|
; GFX900-LABEL: v_shuffle_v2i64_v8i64__14_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v2i64_v8i64__14_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v2i64_v8i64__14_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 15>
|
|
store <2 x i64> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_u() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> poison
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_u() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_u() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s30
|
|
; GFX900-NEXT: s_mov_b32 s9, s31
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s30
|
|
; GFX90A-NEXT: s_mov_b32 s9, s31
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s30
|
|
; GFX900-NEXT: s_mov_b32 s9, s31
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s30
|
|
; GFX90A-NEXT: s_mov_b32 s9, s31
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s30
|
|
; GFX942-NEXT: s_mov_b32 s9, s31
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s30
|
|
; GFX900-NEXT: s_mov_b32 s13, s31
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s30
|
|
; GFX90A-NEXT: s_mov_b32 s13, s31
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s50
|
|
; GFX900-NEXT: s_mov_b32 s9, s51
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s50
|
|
; GFX90A-NEXT: s_mov_b32 s9, s51
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s30
|
|
; GFX942-NEXT: s_mov_b32 s9, s31
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s50
|
|
; GFX900-NEXT: s_mov_b32 s17, s51
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s50
|
|
; GFX90A-NEXT: s_mov_b32 s17, s51
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s30
|
|
; GFX942-NEXT: s_mov_b32 s13, s31
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_9() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__15_9:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s22
|
|
; GFX9-NEXT: s_mov_b32 s9, s23
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_11() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__15_11:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__15_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__15_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__15_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s18
|
|
; GFX900-NEXT: s_mov_b32 s17, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__15_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s18
|
|
; GFX90A-NEXT: s_mov_b32 s17, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__15_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 15, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_0() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> zeroinitializer
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_0() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s4
|
|
; GFX9-NEXT: s_mov_b32 s11, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s4
|
|
; GFX900-NEXT: s_mov_b32 s15, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s4
|
|
; GFX90A-NEXT: s_mov_b32 s15, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s4
|
|
; GFX900-NEXT: s_mov_b32 s19, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s4
|
|
; GFX90A-NEXT: s_mov_b32 s19, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s0
|
|
; GFX942-NEXT: s_mov_b32 s15, s1
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s36
|
|
; GFX900-NEXT: s_mov_b32 s11, s37
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s36
|
|
; GFX90A-NEXT: s_mov_b32 s11, s37
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s4
|
|
; GFX900-NEXT: s_mov_b32 s19, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s4
|
|
; GFX90A-NEXT: s_mov_b32 s19, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_0() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s22, s4
|
|
; GFX900-NEXT: s_mov_b32 s23, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s22, s4
|
|
; GFX90A-NEXT: s_mov_b32 s23, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s18, s0
|
|
; GFX942-NEXT: s_mov_b32 s19, s1
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_1() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_1() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s6
|
|
; GFX9-NEXT: s_mov_b32 s11, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_1() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__3_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s6
|
|
; GFX900-NEXT: s_mov_b32 s15, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s6
|
|
; GFX90A-NEXT: s_mov_b32 s15, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_1() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__5_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s6
|
|
; GFX900-NEXT: s_mov_b32 s19, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s6
|
|
; GFX90A-NEXT: s_mov_b32 s19, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s2
|
|
; GFX942-NEXT: s_mov_b32 s15, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_1() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__7_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s22
|
|
; GFX9-NEXT: s_mov_b32 s9, s23
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s38
|
|
; GFX900-NEXT: s_mov_b32 s11, s39
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s38
|
|
; GFX90A-NEXT: s_mov_b32 s11, s39
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s6
|
|
; GFX900-NEXT: s_mov_b32 s19, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s6
|
|
; GFX90A-NEXT: s_mov_b32 s19, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_1() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s22, s6
|
|
; GFX900-NEXT: s_mov_b32 s23, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s22, s6
|
|
; GFX90A-NEXT: s_mov_b32 s23, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s18, s2
|
|
; GFX942-NEXT: s_mov_b32 s19, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s8
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s8
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_2() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_2() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s8
|
|
; GFX900-NEXT: s_mov_b32 s15, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s8
|
|
; GFX90A-NEXT: s_mov_b32 s15, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s8
|
|
; GFX900-NEXT: s_mov_b32 s19, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s8
|
|
; GFX90A-NEXT: s_mov_b32 s19, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s4
|
|
; GFX942-NEXT: s_mov_b32 s15, s5
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s8
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s8
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s22, s8
|
|
; GFX900-NEXT: s_mov_b32 s23, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s22, s8
|
|
; GFX90A-NEXT: s_mov_b32 s23, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_2() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s26, s8
|
|
; GFX900-NEXT: s_mov_b32 s27, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s26, s8
|
|
; GFX90A-NEXT: s_mov_b32 s27, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s22, s4
|
|
; GFX942-NEXT: s_mov_b32 s23, s5
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_3() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_3() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__1_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_3() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s10
|
|
; GFX900-NEXT: s_mov_b32 s15, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s10
|
|
; GFX90A-NEXT: s_mov_b32 s15, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s6
|
|
; GFX942-NEXT: s_mov_b32 s11, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_3() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__5_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s10
|
|
; GFX900-NEXT: s_mov_b32 s19, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s10
|
|
; GFX90A-NEXT: s_mov_b32 s19, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s6
|
|
; GFX942-NEXT: s_mov_b32 s15, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_3() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__7_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s10, s22
|
|
; GFX900-NEXT: s_mov_b32 s11, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s10, s22
|
|
; GFX90A-NEXT: s_mov_b32 s11, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s22, s10
|
|
; GFX900-NEXT: s_mov_b32 s23, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s22, s10
|
|
; GFX90A-NEXT: s_mov_b32 s23, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_3() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s26, s10
|
|
; GFX900-NEXT: s_mov_b32 s27, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s26, s10
|
|
; GFX90A-NEXT: s_mov_b32 s27, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s22, s6
|
|
; GFX942-NEXT: s_mov_b32 s23, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s8
|
|
; GFX942-NEXT: s_mov_b32 s11, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s16
|
|
; GFX9-NEXT: s_mov_b32 s11, s17
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__1_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__3_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s12
|
|
; GFX900-NEXT: s_mov_b32 s15, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s12
|
|
; GFX90A-NEXT: s_mov_b32 s15, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s8
|
|
; GFX942-NEXT: s_mov_b32 s11, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__5_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s12
|
|
; GFX900-NEXT: s_mov_b32 s19, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s12
|
|
; GFX90A-NEXT: s_mov_b32 s19, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s8
|
|
; GFX942-NEXT: s_mov_b32 s15, s9
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_4() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__7_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s8
|
|
; GFX942-NEXT: s_mov_b32 s11, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s26, s12
|
|
; GFX900-NEXT: s_mov_b32 s27, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s26, s12
|
|
; GFX90A-NEXT: s_mov_b32 s27, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_4() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s30, s12
|
|
; GFX900-NEXT: s_mov_b32 s31, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s30, s12
|
|
; GFX90A-NEXT: s_mov_b32 s31, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s26, s8
|
|
; GFX942-NEXT: s_mov_b32 s27, s9
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_5() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s18
|
|
; GFX9-NEXT: s_mov_b32 s11, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_5() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s14
|
|
; GFX900-NEXT: s_mov_b32 s19, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s14
|
|
; GFX90A-NEXT: s_mov_b32 s19, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s10
|
|
; GFX942-NEXT: s_mov_b32 s15, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s22
|
|
; GFX900-NEXT: s_mov_b32 s11, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s22
|
|
; GFX90A-NEXT: s_mov_b32 s11, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s22
|
|
; GFX900-NEXT: s_mov_b32 s13, s23
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s22
|
|
; GFX90A-NEXT: s_mov_b32 s13, s23
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s26, s14
|
|
; GFX900-NEXT: s_mov_b32 s27, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s26, s14
|
|
; GFX90A-NEXT: s_mov_b32 s27, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s26
|
|
; GFX900-NEXT: s_mov_b32 s13, s27
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s26
|
|
; GFX90A-NEXT: s_mov_b32 s13, s27
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_5() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s30, s14
|
|
; GFX900-NEXT: s_mov_b32 s31, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s30, s14
|
|
; GFX90A-NEXT: s_mov_b32 s31, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s26, s10
|
|
; GFX942-NEXT: s_mov_b32 s27, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_6() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s20
|
|
; GFX9-NEXT: s_mov_b32 s11, s21
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_6() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s16
|
|
; GFX9-NEXT: s_mov_b32 s11, s17
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s16
|
|
; GFX900-NEXT: s_mov_b32 s15, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s16
|
|
; GFX90A-NEXT: s_mov_b32 s15, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s16
|
|
; GFX900-NEXT: s_mov_b32 s19, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s16
|
|
; GFX90A-NEXT: s_mov_b32 s19, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s12
|
|
; GFX942-NEXT: s_mov_b32 s15, s13
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s38
|
|
; GFX900-NEXT: s_mov_b32 s9, s39
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s38
|
|
; GFX90A-NEXT: s_mov_b32 s9, s39
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s42
|
|
; GFX900-NEXT: s_mov_b32 s9, s43
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s42
|
|
; GFX90A-NEXT: s_mov_b32 s9, s43
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s46, s16
|
|
; GFX900-NEXT: s_mov_b32 s47, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s46, s16
|
|
; GFX90A-NEXT: s_mov_b32 s47, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s46
|
|
; GFX900-NEXT: s_mov_b32 s9, s47
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s46
|
|
; GFX90A-NEXT: s_mov_b32 s9, s47
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_6() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s50, s16
|
|
; GFX900-NEXT: s_mov_b32 s51, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s50, s16
|
|
; GFX90A-NEXT: s_mov_b32 s51, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s30, s12
|
|
; GFX942-NEXT: s_mov_b32 s31, s13
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_7() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__0_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s22
|
|
; GFX9-NEXT: s_mov_b32 s11, s23
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s6
|
|
; GFX900-NEXT: s_mov_b32 s17, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s6
|
|
; GFX90A-NEXT: s_mov_b32 s17, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_7() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__2_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s18
|
|
; GFX9-NEXT: s_mov_b32 s11, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s10
|
|
; GFX900-NEXT: s_mov_b32 s17, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s10
|
|
; GFX90A-NEXT: s_mov_b32 s17, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s18
|
|
; GFX900-NEXT: s_mov_b32 s15, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s18
|
|
; GFX90A-NEXT: s_mov_b32 s15, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s14
|
|
; GFX900-NEXT: s_mov_b32 s17, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s14
|
|
; GFX90A-NEXT: s_mov_b32 s17, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s18
|
|
; GFX900-NEXT: s_mov_b32 s17, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s18
|
|
; GFX90A-NEXT: s_mov_b32 s17, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s38
|
|
; GFX900-NEXT: s_mov_b32 s17, s39
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s38
|
|
; GFX90A-NEXT: s_mov_b32 s17, s39
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s18
|
|
; GFX942-NEXT: s_mov_b32 s13, s19
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s22
|
|
; GFX900-NEXT: s_mov_b32 s11, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s22
|
|
; GFX90A-NEXT: s_mov_b32 s11, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s42
|
|
; GFX900-NEXT: s_mov_b32 s17, s43
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s42
|
|
; GFX90A-NEXT: s_mov_b32 s17, s43
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s22
|
|
; GFX942-NEXT: s_mov_b32 s13, s23
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s46, s18
|
|
; GFX900-NEXT: s_mov_b32 s47, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s46, s18
|
|
; GFX90A-NEXT: s_mov_b32 s47, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s46
|
|
; GFX900-NEXT: s_mov_b32 s17, s47
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s46
|
|
; GFX90A-NEXT: s_mov_b32 s17, s47
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s26
|
|
; GFX942-NEXT: s_mov_b32 s13, s27
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_7() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s50, s18
|
|
; GFX900-NEXT: s_mov_b32 s51, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s50, s18
|
|
; GFX90A-NEXT: s_mov_b32 s51, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s30, s14
|
|
; GFX942-NEXT: s_mov_b32 s31, s15
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_8() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__u_8:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 poison, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 0, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 1, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 2, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 3, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 4, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 5, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 6, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 7, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_8() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_8:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> poison, <2 x i32> <i32 8, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_8() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_8:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s4
|
|
; GFX9-NEXT: s_mov_b32 s11, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s4
|
|
; GFX900-NEXT: s_mov_b32 s15, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s4
|
|
; GFX90A-NEXT: s_mov_b32 s15, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_8() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_8:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s4
|
|
; GFX900-NEXT: s_mov_b32 s19, s5
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_8:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s4
|
|
; GFX90A-NEXT: s_mov_b32 s19, s5
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_8:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s0
|
|
; GFX942-NEXT: s_mov_b32 s15, s1
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 8>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s42
|
|
; GFX900-NEXT: s_mov_b32 s9, s43
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s42
|
|
; GFX90A-NEXT: s_mov_b32 s9, s43
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s18
|
|
; GFX900-NEXT: s_mov_b32 s15, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s18
|
|
; GFX90A-NEXT: s_mov_b32 s15, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s38
|
|
; GFX900-NEXT: s_mov_b32 s19, s39
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s38
|
|
; GFX90A-NEXT: s_mov_b32 s19, s39
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s18
|
|
; GFX942-NEXT: s_mov_b32 s15, s19
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s26
|
|
; GFX900-NEXT: s_mov_b32 s9, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s26
|
|
; GFX90A-NEXT: s_mov_b32 s9, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s26
|
|
; GFX942-NEXT: s_mov_b32 s9, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__8_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__8_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__8_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_9() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__9_9:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_9() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_9:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s6
|
|
; GFX9-NEXT: s_mov_b32 s11, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_9() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__11_9:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s6
|
|
; GFX900-NEXT: s_mov_b32 s15, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s6
|
|
; GFX90A-NEXT: s_mov_b32 s15, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_9() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__13_9:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_9() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_9:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s6
|
|
; GFX900-NEXT: s_mov_b32 s19, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_9:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s6
|
|
; GFX90A-NEXT: s_mov_b32 s19, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_9:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s2
|
|
; GFX942-NEXT: s_mov_b32 s15, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 9>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s8
|
|
; GFX900-NEXT: s_mov_b32 s11, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s8
|
|
; GFX90A-NEXT: s_mov_b32 s11, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s20
|
|
; GFX900-NEXT: s_mov_b32 s15, s21
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s20
|
|
; GFX90A-NEXT: s_mov_b32 s15, s21
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s40
|
|
; GFX900-NEXT: s_mov_b32 s19, s41
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s40
|
|
; GFX90A-NEXT: s_mov_b32 s19, s41
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s20
|
|
; GFX942-NEXT: s_mov_b32 s15, s21
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s40
|
|
; GFX900-NEXT: s_mov_b32 s11, s41
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s40
|
|
; GFX90A-NEXT: s_mov_b32 s11, s41
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_10() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_10:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_10() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_10:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s8
|
|
; GFX900-NEXT: s_mov_b32 s15, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s8
|
|
; GFX90A-NEXT: s_mov_b32 s15, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_10() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_10:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s8
|
|
; GFX900-NEXT: s_mov_b32 s19, s9
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_10:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s8
|
|
; GFX90A-NEXT: s_mov_b32 s19, s9
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_10:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s4
|
|
; GFX942-NEXT: s_mov_b32 s15, s5
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 10>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s18
|
|
; GFX900-NEXT: s_mov_b32 s11, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s18
|
|
; GFX90A-NEXT: s_mov_b32 s11, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s38
|
|
; GFX900-NEXT: s_mov_b32 s9, s39
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s38
|
|
; GFX90A-NEXT: s_mov_b32 s9, s39
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s18
|
|
; GFX900-NEXT: s_mov_b32 s11, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s18
|
|
; GFX90A-NEXT: s_mov_b32 s11, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s22
|
|
; GFX900-NEXT: s_mov_b32 s15, s23
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s22
|
|
; GFX90A-NEXT: s_mov_b32 s15, s23
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s18
|
|
; GFX942-NEXT: s_mov_b32 s11, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s42
|
|
; GFX900-NEXT: s_mov_b32 s19, s43
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s42
|
|
; GFX90A-NEXT: s_mov_b32 s19, s43
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s22
|
|
; GFX942-NEXT: s_mov_b32 s15, s23
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s22
|
|
; GFX942-NEXT: s_mov_b32 s9, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_11() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_11:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_11() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__9_11:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__10_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__10_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__10_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_11() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__11_11:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s10
|
|
; GFX900-NEXT: s_mov_b32 s15, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s10
|
|
; GFX90A-NEXT: s_mov_b32 s15, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s6
|
|
; GFX942-NEXT: s_mov_b32 s11, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_11() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__13_11:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_11() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_11:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s10
|
|
; GFX900-NEXT: s_mov_b32 s19, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_11:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s10
|
|
; GFX90A-NEXT: s_mov_b32 s19, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_11:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s6
|
|
; GFX942-NEXT: s_mov_b32 s15, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 11>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s8
|
|
; GFX942-NEXT: s_mov_b32 s11, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s24
|
|
; GFX900-NEXT: s_mov_b32 s15, s25
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s24
|
|
; GFX90A-NEXT: s_mov_b32 s15, s25
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s24
|
|
; GFX900-NEXT: s_mov_b32 s11, s25
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s24
|
|
; GFX90A-NEXT: s_mov_b32 s11, s25
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s44
|
|
; GFX900-NEXT: s_mov_b32 s19, s45
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s44
|
|
; GFX90A-NEXT: s_mov_b32 s19, s45
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s24
|
|
; GFX942-NEXT: s_mov_b32 s15, s25
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s44
|
|
; GFX900-NEXT: s_mov_b32 s11, s45
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s44
|
|
; GFX90A-NEXT: s_mov_b32 s11, s45
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s24
|
|
; GFX942-NEXT: s_mov_b32 s11, s25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s16
|
|
; GFX9-NEXT: s_mov_b32 s11, s17
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__9_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__11_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s12
|
|
; GFX900-NEXT: s_mov_b32 s15, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s12
|
|
; GFX90A-NEXT: s_mov_b32 s15, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s8
|
|
; GFX942-NEXT: s_mov_b32 s11, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_12() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__13_12:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_12() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_12:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s12
|
|
; GFX900-NEXT: s_mov_b32 s19, s13
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_12:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s12
|
|
; GFX90A-NEXT: s_mov_b32 s19, s13
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_12:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s8
|
|
; GFX942-NEXT: s_mov_b32 s15, s9
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 12>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s22
|
|
; GFX900-NEXT: s_mov_b32 s11, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s22
|
|
; GFX90A-NEXT: s_mov_b32 s11, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s6
|
|
; GFX900-NEXT: s_mov_b32 s17, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s6
|
|
; GFX90A-NEXT: s_mov_b32 s17, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s22
|
|
; GFX900-NEXT: s_mov_b32 s11, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s22
|
|
; GFX90A-NEXT: s_mov_b32 s11, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s20, s10
|
|
; GFX900-NEXT: s_mov_b32 s21, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s20, s10
|
|
; GFX90A-NEXT: s_mov_b32 s21, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s26
|
|
; GFX900-NEXT: s_mov_b32 s15, s27
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s26
|
|
; GFX90A-NEXT: s_mov_b32 s15, s27
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s22
|
|
; GFX942-NEXT: s_mov_b32 s11, s23
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s24, s14
|
|
; GFX900-NEXT: s_mov_b32 s25, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s24, s14
|
|
; GFX90A-NEXT: s_mov_b32 s25, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s46
|
|
; GFX900-NEXT: s_mov_b32 s19, s47
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s46
|
|
; GFX90A-NEXT: s_mov_b32 s19, s47
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s26
|
|
; GFX942-NEXT: s_mov_b32 s15, s27
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s44, s18
|
|
; GFX900-NEXT: s_mov_b32 s45, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s44, s18
|
|
; GFX90A-NEXT: s_mov_b32 s45, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[44:45]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[46:47]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_13() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_13:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s18
|
|
; GFX9-NEXT: s_mov_b32 s11, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_13() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_13:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_13() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_13:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s14
|
|
; GFX900-NEXT: s_mov_b32 s19, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_13:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s14
|
|
; GFX90A-NEXT: s_mov_b32 s19, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_13:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s10
|
|
; GFX942-NEXT: s_mov_b32 s15, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 13>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s24
|
|
; GFX900-NEXT: s_mov_b32 s11, s25
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s24
|
|
; GFX90A-NEXT: s_mov_b32 s11, s25
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s24
|
|
; GFX942-NEXT: s_mov_b32 s11, s25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s24
|
|
; GFX900-NEXT: s_mov_b32 s11, s25
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s24
|
|
; GFX90A-NEXT: s_mov_b32 s11, s25
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s24
|
|
; GFX942-NEXT: s_mov_b32 s11, s25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s24
|
|
; GFX900-NEXT: s_mov_b32 s11, s25
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s24
|
|
; GFX90A-NEXT: s_mov_b32 s11, s25
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s20
|
|
; GFX942-NEXT: s_mov_b32 s11, s21
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s28
|
|
; GFX900-NEXT: s_mov_b32 s15, s29
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s28
|
|
; GFX90A-NEXT: s_mov_b32 s15, s29
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s24
|
|
; GFX942-NEXT: s_mov_b32 s11, s25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s28
|
|
; GFX900-NEXT: s_mov_b32 s11, s29
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s28
|
|
; GFX90A-NEXT: s_mov_b32 s11, s29
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s24
|
|
; GFX942-NEXT: s_mov_b32 s11, s25
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s48
|
|
; GFX900-NEXT: s_mov_b32 s19, s49
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s48
|
|
; GFX90A-NEXT: s_mov_b32 s19, s49
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s28
|
|
; GFX942-NEXT: s_mov_b32 s15, s29
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s48
|
|
; GFX900-NEXT: s_mov_b32 s11, s49
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s48
|
|
; GFX90A-NEXT: s_mov_b32 s11, s49
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s28
|
|
; GFX942-NEXT: s_mov_b32 s11, s29
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_14() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_14:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s20
|
|
; GFX9-NEXT: s_mov_b32 s11, s21
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_14() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_14:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s16
|
|
; GFX9-NEXT: s_mov_b32 s11, s17
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s16
|
|
; GFX900-NEXT: s_mov_b32 s15, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s16
|
|
; GFX90A-NEXT: s_mov_b32 s15, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_14() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_14:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s16
|
|
; GFX900-NEXT: s_mov_b32 s19, s17
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_14:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s16
|
|
; GFX90A-NEXT: s_mov_b32 s19, s17
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_14:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s12
|
|
; GFX942-NEXT: s_mov_b32 s15, s13
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 14>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__u_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__u_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__u_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__u_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 poison, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__0_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__0_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s26
|
|
; GFX900-NEXT: s_mov_b32 s11, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__0_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s26
|
|
; GFX90A-NEXT: s_mov_b32 s11, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__0_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s26
|
|
; GFX942-NEXT: s_mov_b32 s11, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 0, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__1_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__1_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s20, s6
|
|
; GFX900-NEXT: s_mov_b32 s21, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__1_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s20, s6
|
|
; GFX90A-NEXT: s_mov_b32 s21, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__1_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s16, s2
|
|
; GFX942-NEXT: s_mov_b32 s17, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 1, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__2_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__2_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s26
|
|
; GFX900-NEXT: s_mov_b32 s11, s27
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__2_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s26
|
|
; GFX90A-NEXT: s_mov_b32 s11, s27
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__2_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s26
|
|
; GFX942-NEXT: s_mov_b32 s11, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 2, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__3_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__3_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:27]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s24, s10
|
|
; GFX900-NEXT: s_mov_b32 s25, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__3_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:27]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s24, s10
|
|
; GFX90A-NEXT: s_mov_b32 s25, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__3_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:23]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s20, s6
|
|
; GFX942-NEXT: s_mov_b32 s21, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[20:21]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[22:23]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 3, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__4_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__4_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s30
|
|
; GFX900-NEXT: s_mov_b32 s15, s31
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__4_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s30
|
|
; GFX90A-NEXT: s_mov_b32 s15, s31
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__4_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s26
|
|
; GFX942-NEXT: s_mov_b32 s11, s27
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 4, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__5_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__5_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:31]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s28, s14
|
|
; GFX900-NEXT: s_mov_b32 s29, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX900-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__5_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:31]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s28, s14
|
|
; GFX90A-NEXT: s_mov_b32 s29, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__5_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:27]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s24, s10
|
|
; GFX942-NEXT: s_mov_b32 s25, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[24:25]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[26:27]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 5, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__6_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__6_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s18, s50
|
|
; GFX900-NEXT: s_mov_b32 s19, s51
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__6_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s18, s50
|
|
; GFX90A-NEXT: s_mov_b32 s19, s51
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__6_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s14, s30
|
|
; GFX942-NEXT: s_mov_b32 s15, s31
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 6, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__7_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__7_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX900-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX900-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX900-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX900-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX900-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[36:51]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s48, s18
|
|
; GFX900-NEXT: s_mov_b32 s49, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX900-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX900-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX900-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX900-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX900-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX900-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX900-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX900-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX900-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__7_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s36, 0
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s37, 1
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s38, 2
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s39, 3
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s48, 4
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s49, 5
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
|
|
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[36:51]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s48, s18
|
|
; GFX90A-NEXT: s_mov_b32 s49, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[48:49]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[50:51]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_readlane_b32 s51, v0, 7
|
|
; GFX90A-NEXT: v_readlane_b32 s50, v0, 6
|
|
; GFX90A-NEXT: v_readlane_b32 s49, v0, 5
|
|
; GFX90A-NEXT: v_readlane_b32 s48, v0, 4
|
|
; GFX90A-NEXT: v_readlane_b32 s39, v0, 3
|
|
; GFX90A-NEXT: v_readlane_b32 s38, v0, 2
|
|
; GFX90A-NEXT: v_readlane_b32 s37, v0, 1
|
|
; GFX90A-NEXT: v_readlane_b32 s36, v0, 0
|
|
; GFX90A-NEXT: s_xor_saveexec_b64 s[4:5], -1
|
|
; GFX90A-NEXT: buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
|
|
; GFX90A-NEXT: s_mov_b64 exec, s[4:5]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__7_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_store_dword off, v0, s32 ; 4-byte Folded Spill
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: v_writelane_b32 v0, s30, 0
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[16:31]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s28, s14
|
|
; GFX942-NEXT: s_mov_b32 s29, s15
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[28:29]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[30:31]
|
|
; GFX942-NEXT: v_readlane_b32 s30, v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_readlane_b32 s31, v0, 1
|
|
; GFX942-NEXT: s_xor_saveexec_b64 s[0:1], -1
|
|
; GFX942-NEXT: scratch_load_dword v0, off, s32 ; 4-byte Folded Reload
|
|
; GFX942-NEXT: s_mov_b64 exec, s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 7, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__8_15() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__8_15:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:23]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s22
|
|
; GFX9-NEXT: s_mov_b32 s11, s23
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 8, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__9_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__9_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s6
|
|
; GFX900-NEXT: s_mov_b32 s17, s7
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__9_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s6
|
|
; GFX90A-NEXT: s_mov_b32 s17, s7
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__9_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 9, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__10_15() #0 {
|
|
; GFX9-LABEL: s_shuffle_v2i64_v8i64__10_15:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s18
|
|
; GFX9-NEXT: s_mov_b32 s11, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 10, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__11_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__11_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s10
|
|
; GFX900-NEXT: s_mov_b32 s17, s11
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__11_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s10
|
|
; GFX90A-NEXT: s_mov_b32 s17, s11
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__11_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 11, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__12_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__12_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s14, s18
|
|
; GFX900-NEXT: s_mov_b32 s15, s19
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__12_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s14, s18
|
|
; GFX90A-NEXT: s_mov_b32 s15, s19
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__12_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 12, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__13_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__13_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s16, s14
|
|
; GFX900-NEXT: s_mov_b32 s17, s15
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__13_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s16, s14
|
|
; GFX90A-NEXT: s_mov_b32 s17, s15
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__13_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 13, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v2i64_v8i64__14_15() #0 {
|
|
; GFX900-LABEL: s_shuffle_v2i64_v8i64__14_15:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX900-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v2i64_v8i64__14_15:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b64 s[8:9], s[16:17]
|
|
; GFX90A-NEXT: s_mov_b64 s[10:11], s[18:19]
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v2i64_v8i64__14_15:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
|
|
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <8 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <8 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <8 x i64> %vec0, <8 x i64> %vec1, <2 x i32> <i32 14, i32 15>
|
|
call void asm sideeffect "; use $0", "{s[8:11]}"(<2 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind }
|
|
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GFX90APLUS: {{.*}}
|