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
3736 lines
153 KiB
LLVM
3736 lines
153 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck --check-prefix=GFX9 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji < %s | FileCheck --check-prefix=GFX8 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=hawaii < %s | FileCheck --check-prefix=GFX7 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck --check-prefix=GFX10 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=+real-true16 < %s | FileCheck --check-prefixes=GFX11,GFX11-TRUE16 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -mattr=-real-true16 < %s | FileCheck --check-prefixes=GFX11,GFX11-FAKE16 %s
|
|
|
|
define amdgpu_ps void @insertelement_s_v2i8_s_s(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: global_load_ushort v0, v0, s[2:3]
|
|
; GFX9-NEXT: s_mov_b32 m0, s5
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX9-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX9-NEXT: s_movreld_b32 s0, s4
|
|
; GFX9-NEXT: s_and_b32 s1, s1, 0xff
|
|
; GFX9-NEXT: s_and_b32 s0, s0, 0xff
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 8
|
|
; GFX9-NEXT: s_or_b32 s0, s0, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX9-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX8-NEXT: flat_load_ushort v0, v[0:1]
|
|
; GFX8-NEXT: s_mov_b32 m0, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX8-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX8-NEXT: s_movreld_b32 s0, s4
|
|
; GFX8-NEXT: s_and_b32 s1, s1, 0xff
|
|
; GFX8-NEXT: s_and_b32 s0, s0, 0xff
|
|
; GFX8-NEXT: s_lshl_b32 s1, s1, 8
|
|
; GFX8-NEXT: s_or_b32 s0, s0, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s0, s2
|
|
; GFX7-NEXT: s_mov_b32 s1, s3
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_load_ushort v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_mov_b32 m0, s5
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_readfirstlane_b32 s6, v0
|
|
; GFX7-NEXT: s_lshr_b32 s7, s6, 8
|
|
; GFX7-NEXT: s_movreld_b32 s6, s4
|
|
; GFX7-NEXT: s_and_b32 s5, s7, 0xff
|
|
; GFX7-NEXT: s_and_b32 s4, s6, 0xff
|
|
; GFX7-NEXT: s_lshl_b32 s5, s5, 8
|
|
; GFX7-NEXT: s_or_b32 s4, s4, s5
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_mov_b32 m0, s5
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: global_load_ushort v0, v0, s[2:3]
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX10-NEXT: s_movreld_b32 s0, s4
|
|
; GFX10-NEXT: s_and_b32 s1, s1, 0xff
|
|
; GFX10-NEXT: s_and_b32 s0, s0, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s1, s1, 8
|
|
; GFX10-NEXT: s_or_b32 s0, s0, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX10-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
|
|
; GFX11-TRUE16-NEXT: s_mov_b32 m0, s5
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v0, v0, s[2:3]
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
|
|
; GFX11-TRUE16-NEXT: s_movreld_b32 s0, s4
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s1, s1, 0xff
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s0, s0, 0xff
|
|
; GFX11-TRUE16-NEXT: s_lshl_b32 s1, s1, 8
|
|
; GFX11-TRUE16-NEXT: s_or_b32 s0, s0, s1
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: v_mov_b16_e32 v0.l, s0
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_s_v2i8_s_s:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_mov_b32 m0, s5
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v0, v0, s[2:3]
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX11-FAKE16-NEXT: s_movreld_b32 s0, s4
|
|
; GFX11-FAKE16-NEXT: s_and_b32 s1, s1, 0xff
|
|
; GFX11-FAKE16-NEXT: s_and_b32 s0, s0, 0xff
|
|
; GFX11-FAKE16-NEXT: s_lshl_b32 s1, s1, 8
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-FAKE16-NEXT: s_or_b32 s0, s0, s1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v2i8_s_s(ptr addrspace(1) %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s3, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v3, 8, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[1:2], v0, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_ushort v0, v[0:1]
|
|
; GFX8-NEXT: s_mov_b32 m0, s3
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX8-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 8, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[1:2], v0
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_ushort v0, v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_mov_b32 m0, s3
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0xff
|
|
; GFX10-NEXT: s_mov_b32 m0, s3
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, s2
|
|
; GFX10-NEXT: v_and_b32_sdwa v3, v1, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[1:2], v0, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v1, v[0:1], off
|
|
; GFX11-TRUE16-NEXT: s_mov_b32 m0, s3
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_lshrrev_b16 v2.l, 8, v1.l
|
|
; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v1, s2
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_v_v2i8_s_s:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-FAKE16-NEXT: s_mov_b32 m0, s3
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v0, s2
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(1 ) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v2i8_v_s(ptr addrspace(4) inreg %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: global_load_ushort v1, v1, s[2:3]
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX9-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s4, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v2
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v0, 8, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[2:3], v0, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s3
|
|
; GFX8-NEXT: flat_load_ushort v1, v[1:2]
|
|
; GFX8-NEXT: s_mov_b32 m0, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX8-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v1, v0
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v2
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v0, 8, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[2:3], v0
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s0, s2
|
|
; GFX7-NEXT: s_mov_b32 s1, s3
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_load_ushort v1, off, s[0:3], 0
|
|
; GFX7-NEXT: s_mov_b32 m0, s4
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_readfirstlane_b32 s4, v1
|
|
; GFX7-NEXT: s_lshr_b32 s5, s4, 8
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s5
|
|
; GFX7-NEXT: v_movreld_b32_e32 v1, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_mov_b32 m0, s4
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: global_load_ushort v1, v1, s[2:3]
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX10-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v1, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0xff
|
|
; GFX10-NEXT: v_and_b32_sdwa v0, v2, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[2:3], v0, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: s_mov_b32 m0, s4
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v1, v1, s[2:3]
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
|
|
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v1, v0
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_s_v2i8_v_s:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_mov_b32 m0, s4
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v1, v1, s[2:3]
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v1, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v0, 0xff, v2
|
|
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v2, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v2i8_s_v(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: global_load_ushort v1, v1, s[2:3]
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX9-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v0, v3, v2, vcc
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s3
|
|
; GFX8-NEXT: flat_load_ushort v1, v[1:2]
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX8-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v0, v3, v2, vcc
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s0, s2
|
|
; GFX7-NEXT: s_mov_b32 s1, s3
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_load_ushort v1, off, s[0:3], 0
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_readfirstlane_b32 s4, v1
|
|
; GFX7-NEXT: s_lshr_b32 s5, s4, 8
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v0, v3, v2, vcc
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 8, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0xff
|
|
; GFX10-NEXT: global_load_ushort v1, v1, s[2:3]
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX10-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_and_b32_sdwa v3, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v1, v1, s[2:3]
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v2, s4, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_s_v2i8_s_v:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v1, v1, s[2:3]
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v1
|
|
; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s4, vcc_lo
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v2, s4, vcc_lo
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v2i8_v_v(ptr addrspace(4) inreg %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: global_load_ushort v2, v2, s[2:3]
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX9-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v0, v3, v0, vcc
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v3, 8, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: flat_load_ushort v2, v[2:3]
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX8-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v0, v3, v0, vcc
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 8, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s0, s2
|
|
; GFX7-NEXT: s_mov_b32 s1, s3
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_load_ushort v2, off, s[0:3], 0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_readfirstlane_b32 s4, v2
|
|
; GFX7-NEXT: s_lshr_b32 s5, s4, 8
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v2, v2, v0, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v0, v3, v0, vcc
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 8, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0xff
|
|
; GFX10-NEXT: global_load_ushort v2, v2, s[2:3]
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX10-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_and_b32_sdwa v2, v2, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v4, s0, v0, vcc_lo
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v2, v2, s[2:3]
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX11-TRUE16-NEXT: s_and_b32 s1, 0xffff, s0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-TRUE16-NEXT: s_lshr_b32 s1, s1, 8
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, s0, v0, vcc_lo
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_s_v2i8_v_v:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v2, v2, s[2:3]
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX11-FAKE16-NEXT: s_lshr_b32 s1, s0, 8
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v2, s1, v0, vcc_lo
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v1
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v2
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, s0, v0, vcc_lo
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v2i8_s_v(ptr addrspace(1) %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 8, v0
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v2
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v0, v3, v1, vcc
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_ushort v0, v[0:1]
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 8, v0
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v2
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v0, v3, v1, vcc
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_ushort v0, v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s2
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 0, v2
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 8, v0
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v1, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v2
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0xff
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
|
|
; GFX10-NEXT: v_and_b32_sdwa v3, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v0, s2, vcc_lo
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e64 v2, v0, s2, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_v_v2i8_s_v:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v2
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v1, v1, s2, vcc_lo
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v2
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e64 v0, v0, s2, vcc_lo
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_4)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v2i8_v_s(ptr addrspace(1) %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s2, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v2
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v3, 8, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[1:2], v0, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_ushort v0, v[0:1]
|
|
; GFX8-NEXT: s_mov_b32 m0, s2
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX8-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v3, 8, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[1:2], v0
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 m0, s2
|
|
; GFX7-NEXT: s_mov_b32 s2, 0
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: buffer_load_ushort v0, v[0:1], s[0:3], 0 addr64
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX10-NEXT: s_mov_b32 m0, s2
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0xff
|
|
; GFX10-NEXT: v_and_b32_sdwa v3, v1, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[1:2], v0, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v3, v[0:1], off
|
|
; GFX11-TRUE16-NEXT: s_mov_b32 m0, s2
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_lshrrev_b16 v4.l, 8, v3.l
|
|
; GFX11-TRUE16-NEXT: v_movreld_b32_e32 v3, v2
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v4.l
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v3.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_v_v2i8_v_s:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-FAKE16-NEXT: s_mov_b32 m0, s2
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX11-FAKE16-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v2i8_v_v(ptr addrspace(1) %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v4, v0, v2, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v0, v1, v2, vcc
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX9-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_ushort v0, v[0:1]
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v4, v0, v2, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v0, v1, v2, vcc
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX8-NEXT: v_lshlrev_b16_e32 v2, 8, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX8-NEXT: flat_store_short v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s2, 0
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: buffer_load_ushort v0, v[0:1], s[0:3], 0 addr64
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 0, v3
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 8, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_short v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_ushort v0, v[0:1], off
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0xff
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
|
|
; GFX10-NEXT: v_and_b32_sdwa v3, v1, v4 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_or_b32_sdwa v2, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
|
|
; GFX10-NEXT: global_store_short v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-TRUE16-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX11-TRUE16: ; %bb.0:
|
|
; GFX11-TRUE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
|
|
; GFX11-TRUE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-TRUE16-NEXT: v_lshrrev_b16 v1.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
|
|
; GFX11-TRUE16-NEXT: v_cndmask_b32_e32 v2, v0, v2, vcc_lo
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.h, 0xff, v2.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-TRUE16-NEXT: v_and_b16 v0.l, 0xff, v1.l
|
|
; GFX11-TRUE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-TRUE16-NEXT: v_lshlrev_b16 v0.l, 8, v0.l
|
|
; GFX11-TRUE16-NEXT: v_or_b16 v0.l, v0.h, v0.l
|
|
; GFX11-TRUE16-NEXT: global_store_b16 v[1:2], v0, off
|
|
; GFX11-TRUE16-NEXT: s_endpgm
|
|
;
|
|
; GFX11-FAKE16-LABEL: insertelement_v_v2i8_v_v:
|
|
; GFX11-FAKE16: ; %bb.0:
|
|
; GFX11-FAKE16-NEXT: global_load_u16 v0, v[0:1], off
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3
|
|
; GFX11-FAKE16-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-FAKE16-NEXT: v_lshrrev_b32_e32 v1, 8, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc_lo
|
|
; GFX11-FAKE16-NEXT: v_cmp_eq_u32_e32 vcc_lo, 0, v3
|
|
; GFX11-FAKE16-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc_lo
|
|
; GFX11-FAKE16-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-FAKE16-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_and_b32 v1, 0xff, v1
|
|
; GFX11-FAKE16-NEXT: v_lshlrev_b16 v3, 8, v1
|
|
; GFX11-FAKE16-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-FAKE16-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX11-FAKE16-NEXT: global_store_b16 v[0:1], v2, off
|
|
; GFX11-FAKE16-NEXT: s_endpgm
|
|
%vec = load <2 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <2 x i8> %vec, i8 %val, i32 %idx
|
|
store <2 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
; FIXME: 3 element load/store legalization
|
|
; define amdgpu_ps void @insertelement_s_v3i8_s_s(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(4) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_v_v3i8_s_s(ptr addrspace(1) %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(1 ) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_s_v3i8_v_s(ptr addrspace(4) inreg %ptr, i8 %val, i32 inreg %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(4) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_s_v3i8_s_v(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(4) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_s_v3i8_v_v(ptr addrspace(4) inreg %ptr, i8 %val, i32 %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(4) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_v_v3i8_s_v(ptr addrspace(1) %ptr, i8 inreg %val, i32 %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(1) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_v_v3i8_v_s(ptr addrspace(1) %ptr, i8 %val, i32 inreg %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(1) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
; define amdgpu_ps void @insertelement_v_v3i8_v_v(ptr addrspace(1) %ptr, i8 %val, i32 %idx) {
|
|
; %vec = load <3 x i8>, ptr addrspace(1) %ptr
|
|
; %insert = insertelement <3 x i8> %vec, i8 %val, i32 %idx
|
|
; store <3 x i8> %insert, ptr addrspace(1) null
|
|
; ret void
|
|
; }
|
|
|
|
define amdgpu_ps void @insertelement_v_v4i8_s_s(ptr addrspace(1) %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v4i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dword v2, v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX9-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX9-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX9-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX9-NEXT: s_not_b32 s0, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, v2, s0, v3
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v4i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dword v0, v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX8-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v2, s0, 0, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, s1, v2
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v4i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX7-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v0, s0, 0, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, s1, v0
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v4i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dword v2, v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX10-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX10-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX10-NEXT: s_not_b32 s1, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, v2, s1, s0
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v4i8_s_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b32 v2, v[0:1], off
|
|
; GFX11-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX11-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
|
|
; GFX11-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX11-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX11-NEXT: s_not_b32 s1, s2
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_and_or_b32 v2, v2, s1, s0
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(1 ) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v4i8_v_s(ptr addrspace(4) inreg %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v4i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX9-NEXT: s_and_b32 s1, s4, 3
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX9-NEXT: s_lshl_b32 s2, 0xff, s1
|
|
; GFX9-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: s_andn2_b32 s0, s0, s2
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_lshl_or_b32 v2, v2, s1, v3
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v4i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX8-NEXT: s_and_b32 s1, s4, 3
|
|
; GFX8-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_andn2_b32 s0, s0, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, s0, v2
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v4i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX7-NEXT: s_and_b32 s1, s4, 3
|
|
; GFX7-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, s1, v0
|
|
; GFX7-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_andn2_b32 s0, s0, s1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, s0, v0
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v4i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX10-NEXT: s_and_b32 s1, s4, 3
|
|
; GFX10-NEXT: v_and_b32_e32 v2, 0xff, v0
|
|
; GFX10-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_lshl_b32 s2, 0xff, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: s_andn2_b32 s0, s0, s2
|
|
; GFX10-NEXT: v_lshl_or_b32 v2, v2, s1, s0
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v4i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
|
|
; GFX11-NEXT: s_and_b32 s1, s4, 3
|
|
; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v2, 0xff, v0
|
|
; GFX11-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-NEXT: s_lshl_b32 s2, 0xff, s1
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_and_not1_b32 s0, s0, s2
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_lshl_or_b32 v2, v2, s1, s0
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v4i8_s_v(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v4i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v2, v0, s1
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v0, v1
|
|
; GFX9-NEXT: v_not_b32_e32 v3, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, s0, v3, v2
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v4i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v2, v0, s1
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v0, v1
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v3, v0, 0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v4i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, s1, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v0, v0, 0, s0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v4i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v1, v0, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v0, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, s0, v3, v2
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v4i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX11-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v1, v0, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v0, s1
|
|
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_not_b32_e32 v3, v1
|
|
; GFX11-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: v_and_or_b32 v2, s0, v3, v2
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v4i8_v_v(ptr addrspace(4) inreg %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v4i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX9-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v3, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v1, v2
|
|
; GFX9-NEXT: v_not_b32_e32 v2, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, s0, v2, v3
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v4i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX8-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v3, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v1, v2
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v2, v0, 0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v4i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v1, v1, 0, s0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v4i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dword s0, s[2:3], 0x0
|
|
; GFX10-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v1, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v3, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_not_b32_e32 v2, v2
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, s0, v2, v3
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v4i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b32 s0, s[2:3], 0x0
|
|
; GFX11-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v1, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_lshlrev_b32 v3, v1, v0
|
|
; GFX11-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-NEXT: v_not_b32_e32 v2, v2
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_and_or_b32 v2, s0, v2, v3
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v4i8_s_v(ptr addrspace(1) %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v4i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dword v3, v[0:1], off
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 3, v2
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v2, v0, s0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v0, v1
|
|
; GFX9-NEXT: v_not_b32_e32 v4, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, v3, v4, v2
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v4i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dword v0, v[0:1]
|
|
; GFX8-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, v2, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v3, v2, s0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v2, v1, 0, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v2, v3
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v4i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v2
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_lshl_b32_e32 v2, s0, v1
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v0, v1, 0, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v2
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v4i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dword v3, v[0:1], off
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 3, v2
|
|
; GFX10-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v1, v0, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_not_b32_e32 v4, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, v3, v4, v2
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v4i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b32 v3, v[0:1], off
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 3, v2
|
|
; GFX11-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v1, v0, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v0, s0
|
|
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_not_b32_e32 v4, v1
|
|
; GFX11-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_and_or_b32 v2, v3, v4, v2
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v4i8_v_s(ptr addrspace(1) %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v4i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dword v3, v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX9-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX9-NEXT: s_not_b32 s0, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, v3, s0, v2
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v4i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dword v0, v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v3, s0, 0, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v4i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 0xff, v2
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, s0, v1
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v0, s0, 0, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v1
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v4i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dword v3, v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX10-NEXT: s_not_b32 s0, s0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, v3, s0, v2
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v4i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b32 v3, v[0:1], off
|
|
; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_and_b32 v0, 0xff, v2
|
|
; GFX11-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, s0, v0
|
|
; GFX11-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX11-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX11-NEXT: s_not_b32 s0, s0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_and_or_b32 v2, v3, s0, v2
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v4i8_v_v(ptr addrspace(1) %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v4i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dword v4, v[0:1], off
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 3, v3
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v0, v1
|
|
; GFX9-NEXT: v_not_b32_e32 v3, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_and_or_b32 v2, v4, v3, v2
|
|
; GFX9-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v4i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dword v0, v[0:1]
|
|
; GFX8-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, v3, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_bfi_b32 v3, v1, 0, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX8-NEXT: flat_store_dword v[0:1], v2
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v4i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s2, 0
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: buffer_load_dword v0, v[0:1], s[0:3], 0 addr64
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v3
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v2, v1, v2
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_bfi_b32 v0, v1, 0, v0
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v0, v2
|
|
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v4i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dword v4, v[0:1], off
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 3, v3
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v1, v0, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_and_or_b32 v2, v4, v3, v2
|
|
; GFX10-NEXT: global_store_dword v[0:1], v2, off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v4i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b32 v4, v[0:1], off
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 3, v3
|
|
; GFX11-NEXT: v_and_b32_e32 v1, 0xff, v2
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v0, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_lshlrev_b32 v3, v0, v1
|
|
; GFX11-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX11-NEXT: v_not_b32_e32 v2, v2
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_and_or_b32 v2, v4, v2, v3
|
|
; GFX11-NEXT: global_store_b32 v[0:1], v2, off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <4 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <4 x i8> %vec, i8 %val, i32 %idx
|
|
store <4 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v8i8_s_s(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v8i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX9-NEXT: s_and_b32 s3, s5, 3
|
|
; GFX9-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX9-NEXT: s_lshl_b32 s3, s3, 3
|
|
; GFX9-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: s_movrels_b32 s2, s0
|
|
; GFX9-NEXT: s_lshl_b32 s4, s4, s3
|
|
; GFX9-NEXT: s_lshl_b32 s3, 0xff, s3
|
|
; GFX9-NEXT: s_andn2_b32 s2, s2, s3
|
|
; GFX9-NEXT: s_or_b32 s2, s2, s4
|
|
; GFX9-NEXT: s_movreld_b32 s0, s2
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX9-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v8i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX8-NEXT: s_and_b32 s3, s5, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX8-NEXT: s_lshl_b32 s3, s3, 3
|
|
; GFX8-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_movrels_b32 s2, s0
|
|
; GFX8-NEXT: s_lshl_b32 s4, s4, s3
|
|
; GFX8-NEXT: s_lshl_b32 s3, 0xff, s3
|
|
; GFX8-NEXT: s_andn2_b32 s2, s2, s3
|
|
; GFX8-NEXT: s_or_b32 s2, s2, s4
|
|
; GFX8-NEXT: s_movreld_b32 s0, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v8i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx2 s[2:3], s[2:3], 0x0
|
|
; GFX7-NEXT: s_and_b32 s1, s5, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX7-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX7-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_movrels_b32 s0, s2
|
|
; GFX7-NEXT: s_lshl_b32 s4, s4, s1
|
|
; GFX7-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX7-NEXT: s_andn2_b32 s0, s0, s1
|
|
; GFX7-NEXT: s_or_b32 s0, s0, s4
|
|
; GFX7-NEXT: s_movreld_b32 s2, s0
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s2
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v8i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX10-NEXT: s_and_b32 s2, s5, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX10-NEXT: s_lshl_b32 s2, s2, 3
|
|
; GFX10-NEXT: s_and_b32 s3, s4, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s5, 0xff, s2
|
|
; GFX10-NEXT: s_lshl_b32 s2, s3, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: s_movrels_b32 s4, s0
|
|
; GFX10-NEXT: s_andn2_b32 s3, s4, s5
|
|
; GFX10-NEXT: s_or_b32 s2, s3, s2
|
|
; GFX10-NEXT: s_movreld_b32 s0, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v8i8_s_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
|
|
; GFX11-NEXT: s_and_b32 s2, s5, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX11-NEXT: s_lshl_b32 s2, s2, 3
|
|
; GFX11-NEXT: s_and_b32 s3, s4, 0xff
|
|
; GFX11-NEXT: s_lshl_b32 s5, 0xff, s2
|
|
; GFX11-NEXT: s_lshl_b32 s2, s3, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_movrels_b32 s4, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_and_not1_b32 s3, s4, s5
|
|
; GFX11-NEXT: s_or_b32 s2, s3, s2
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_movreld_b32 s0, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v3, s1 :: v_dual_mov_b32 v2, s0
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v8i8_s_s(ptr addrspace(1) %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v8i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s1, s3, 3
|
|
; GFX9-NEXT: s_and_b32 s2, s2, 0xff
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX9-NEXT: s_lshr_b32 s0, s3, 2
|
|
; GFX9-NEXT: s_lshl_b32 s2, s2, s1
|
|
; GFX9-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX9-NEXT: s_not_b32 s1, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_or_b32 v4, v3, s1, v2
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v8i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX8-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_movrels_b32_e32 v2, v0
|
|
; GFX8-NEXT: v_bfi_b32 v2, s0, 0, v2
|
|
; GFX8-NEXT: v_or_b32_e32 v4, s1, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX8-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v8i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX7-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_movrels_b32_e32 v2, v0
|
|
; GFX7-NEXT: v_bfi_b32 v2, s0, 0, v2
|
|
; GFX7-NEXT: v_or_b32_e32 v2, s1, v2
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v8i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX10-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX10-NEXT: s_not_b32 s1, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_movrels_b32_e32 v2, v0
|
|
; GFX10-NEXT: v_and_or_b32 v4, v2, s1, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v8i8_s_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
|
|
; GFX11-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX11-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX11-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX11-NEXT: s_not_b32 s1, s2
|
|
; GFX11-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_movrels_b32_e32 v2, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_and_or_b32 v4, v2, s1, s0
|
|
; GFX11-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX11-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(1 ) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v8i8_v_s(ptr addrspace(4) inreg %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v8i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX9-NEXT: s_lshr_b32 s2, s4, 2
|
|
; GFX9-NEXT: s_and_b32 s4, s4, 3
|
|
; GFX9-NEXT: s_mov_b32 m0, s2
|
|
; GFX9-NEXT: s_lshl_b32 s4, s4, 3
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: s_movrels_b32 s3, s0
|
|
; GFX9-NEXT: s_lshl_b32 s5, 0xff, s4
|
|
; GFX9-NEXT: s_andn2_b32 s3, s3, s5
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX9-NEXT: v_lshl_or_b32 v4, v0, s4, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s2, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v8i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX8-NEXT: s_and_b32 s3, s4, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX8-NEXT: s_lshl_b32 s3, s3, 3
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s3
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_movrels_b32 s2, s0
|
|
; GFX8-NEXT: s_lshl_b32 s3, 0xff, s3
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_andn2_b32 s2, s2, s3
|
|
; GFX8-NEXT: v_or_b32_e32 v4, s2, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX8-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v8i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX7-NEXT: s_and_b32 s3, s4, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX7-NEXT: s_lshl_b32 s3, s3, 3
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_movrels_b32 s2, s0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, s3, v0
|
|
; GFX7-NEXT: s_lshl_b32 s3, 0xff, s3
|
|
; GFX7-NEXT: s_andn2_b32 s2, s2, s3
|
|
; GFX7-NEXT: v_or_b32_e32 v2, s2, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v8i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX10-NEXT: s_and_b32 s2, s4, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX10-NEXT: s_lshl_b32 s2, s2, 3
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX10-NEXT: s_lshl_b32 s3, 0xff, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: s_movrels_b32 s4, s0
|
|
; GFX10-NEXT: s_andn2_b32 s3, s4, s3
|
|
; GFX10-NEXT: v_lshl_or_b32 v4, v0, s2, s3
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v8i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
|
|
; GFX11-NEXT: s_and_b32 s2, s4, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX11-NEXT: s_lshl_b32 s2, s2, 3
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX11-NEXT: s_lshl_b32 s3, 0xff, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_movrels_b32 s4, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_and_not1_b32 s3, s4, s3
|
|
; GFX11-NEXT: v_lshl_or_b32 v4, v0, s2, s3
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v8i8_s_v(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v8i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v3, 2, v0
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX9-NEXT: s_and_b32 s0, s4, 0xff
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v5, v0, s0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v0, v6
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc
|
|
; GFX9-NEXT: v_not_b32_e32 v0, v0
|
|
; GFX9-NEXT: v_and_or_b32 v5, v4, v0, v5
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v3
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v1, v5, s[0:1]
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v5, vcc
|
|
; GFX9-NEXT: global_store_dwordx2 v[3:4], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v8i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v3, 2, v0
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX8-NEXT: s_and_b32 s0, s4, 0xff
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v5, v0, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v0, v6
|
|
; GFX8-NEXT: v_bfi_b32 v0, v0, 0, v4
|
|
; GFX8-NEXT: v_or_b32_e32 v5, v0, v5
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v3
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v1, v5, s[0:1]
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v2, v5, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[3:4], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v8i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 2, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX7-NEXT: s_and_b32 s0, s4, 0xff
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc
|
|
; GFX7-NEXT: v_lshl_b32_e32 v5, s0, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_bfi_b32 v0, v0, 0, v4
|
|
; GFX7-NEXT: v_or_b32_e32 v4, v0, v5
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v3
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v1, v4, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v2, v4, vcc
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v8i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX10-NEXT: v_and_b32_e32 v1, 3, v0
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v4, 2, v0
|
|
; GFX10-NEXT: s_and_b32 s2, s4, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v2, 3, v1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v4
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v3, v2, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v2, s2
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v3
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v4
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
|
|
; GFX10-NEXT: v_and_or_b32 v5, v5, v3, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v5, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v8i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
|
|
; GFX11-NEXT: v_and_b32_e32 v1, 3, v0
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v4, 2, v0
|
|
; GFX11-NEXT: s_and_b32 s2, s4, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, 3, v1
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v4
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v3, v2, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v2, s2
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 0, v4
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_not_b32_e32 v3, v3
|
|
; GFX11-NEXT: v_and_or_b32 v5, v5, v3, v2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, v5, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v5, vcc_lo
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v8i8_v_v(ptr addrspace(4) inreg %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v8i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v4, 2, v1
|
|
; GFX9-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v4
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v1, v1, v6
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v5, v2, v3, vcc
|
|
; GFX9-NEXT: v_not_b32_e32 v1, v1
|
|
; GFX9-NEXT: v_and_or_b32 v1, v5, v1, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v4
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v2, v1, s[0:1]
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
|
|
; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v8i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v4, 2, v1
|
|
; GFX8-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v4
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v5, v2, v3, vcc
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, v1, v6
|
|
; GFX8-NEXT: v_bfi_b32 v1, v1, 0, v5
|
|
; GFX8-NEXT: v_or_b32_e32 v1, v1, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v4
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v2, v1, s[0:1]
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v8i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v4, 2, v1
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v4
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v5, v2, v3, vcc
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_bfi_b32 v1, v1, 0, v5
|
|
; GFX7-NEXT: v_or_b32_e32 v1, v1, v0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v4
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v2, v1, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v8i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0x0
|
|
; GFX10-NEXT: v_and_b32_e32 v2, 3, v1
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 2, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 3, v2
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v5
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v4, v3, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v4
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v5
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v4, v1, v2, vcc_lo
|
|
; GFX10-NEXT: v_and_or_b32 v6, v4, v3, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, v6, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v6, vcc_lo
|
|
; GFX10-NEXT: global_store_dwordx2 v[3:4], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v8i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b64 s[0:1], s[2:3], 0x0
|
|
; GFX11-NEXT: v_and_b32_e32 v2, 3, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v3, 0xff, v0
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v5, 2, v1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v5
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 0, v5
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v4, v2, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, v2, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_not_b32_e32 v3, v4
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc_lo
|
|
; GFX11-NEXT: v_and_or_b32 v4, v4, v3, v2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v8i8_s_v(ptr addrspace(1) %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v8i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v6, 2, v2
|
|
; GFX9-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0xff
|
|
; GFX9-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v7, v2, s0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v2, v2, v5
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
|
|
; GFX9-NEXT: v_not_b32_e32 v2, v2
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v6
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc
|
|
; GFX9-NEXT: v_and_or_b32 v2, v5, v2, v7
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX9-NEXT: global_store_dwordx2 v[3:4], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v8i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v6, 2, v2
|
|
; GFX8-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0xff
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v7, v2, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v2, v2, v5
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v6
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc
|
|
; GFX8-NEXT: v_bfi_b32 v2, v2, 0, v5
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v2, v7
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[3:4], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v8i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 2, v2
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v3
|
|
; GFX7-NEXT: v_lshl_b32_e32 v4, s0, v2
|
|
; GFX7-NEXT: v_lshl_b32_e32 v2, 0xff, v2
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v3
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc
|
|
; GFX7-NEXT: v_bfi_b32 v2, v2, 0, v5
|
|
; GFX7-NEXT: v_or_b32_e32 v2, v2, v4
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v8i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX10-NEXT: v_and_b32_e32 v3, 3, v2
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v5, 2, v2
|
|
; GFX10-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v5
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v4, v3, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v2, v3, s0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v5
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v4
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc_lo
|
|
; GFX10-NEXT: v_and_or_b32 v4, v4, v3, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v8i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v5, 2, v2
|
|
; GFX11-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v5
|
|
; GFX11-NEXT: v_and_b32_e32 v3, 3, v2
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v4, v3, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v2, v3, s0
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 0, v5
|
|
; GFX11-NEXT: v_not_b32_e32 v3, v4
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc_lo
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_and_or_b32 v4, v4, v3, v2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v8i8_v_s(ptr addrspace(1) %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v8i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s1, s2, 3
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX9-NEXT: s_lshr_b32 s0, s2, 2
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, s1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX9-NEXT: s_not_b32 s1, s1
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_or_b32 v4, v3, s1, v2
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v4
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v8i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_movrels_b32_e32 v3, v0
|
|
; GFX8-NEXT: v_bfi_b32 v3, s0, 0, v3
|
|
; GFX8-NEXT: v_or_b32_e32 v4, v3, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX8-NEXT: flat_store_dwordx2 v[2:3], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v8i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v2, s0, v2
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_movrels_b32_e32 v3, v0
|
|
; GFX7-NEXT: v_bfi_b32 v3, s0, 0, v3
|
|
; GFX7-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v2
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v8i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX10-NEXT: s_not_b32 s0, s0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_movrels_b32_e32 v3, v0
|
|
; GFX10-NEXT: v_and_or_b32 v4, v3, s0, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v8i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
|
|
; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX11-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, s0, v2
|
|
; GFX11-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: s_not_b32 s0, s0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_movrels_b32_e32 v3, v0
|
|
; GFX11-NEXT: v_and_or_b32 v4, v3, s0, v2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v8i8_v_v(ptr addrspace(1) %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v8i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v7, 2, v3
|
|
; GFX9-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX9-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v3, v3, v6
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v7
|
|
; GFX9-NEXT: v_not_b32_e32 v3, v3
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v7
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v6, v0, v1, vcc
|
|
; GFX9-NEXT: v_and_or_b32 v2, v6, v3, v2
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v8i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx2 v[0:1], v[0:1]
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v7, 2, v3
|
|
; GFX8-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v7
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v3, v3, v6
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v7
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v6, v0, v1, vcc
|
|
; GFX8-NEXT: v_bfi_b32 v3, v3, 0, v6
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v8i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v4, 2, v3
|
|
; GFX7-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v4
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v2, v3, v2
|
|
; GFX7-NEXT: v_lshl_b32_e32 v3, 0xff, v3
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 0, v4
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc
|
|
; GFX7-NEXT: v_bfi_b32 v3, v3, 0, v5
|
|
; GFX7-NEXT: v_or_b32_e32 v2, v3, v2
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v0, v2, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v1, v2, vcc
|
|
; GFX7-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v8i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx2 v[0:1], v[0:1], off
|
|
; GFX10-NEXT: v_and_b32_e32 v4, 3, v3
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 2, v3
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v4
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v6
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v5, v4, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_not_b32_e32 v3, v5
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc_lo
|
|
; GFX10-NEXT: v_and_or_b32 v4, v4, v3, v2
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
|
|
; GFX10-NEXT: global_store_dwordx2 v[2:3], v[0:1], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v8i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b64 v[0:1], v[0:1], off
|
|
; GFX11-NEXT: v_and_b32_e32 v4, 3, v3
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v6, 2, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v4, 3, v4
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
|
|
; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 0, v6
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v5, v4, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v2, v4, v2
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_not_b32_e32 v3, v5
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v4, v0, v1, vcc_lo
|
|
; GFX11-NEXT: v_and_or_b32 v4, v4, v3, v2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, v4, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v4, vcc_lo
|
|
; GFX11-NEXT: global_store_b64 v[2:3], v[0:1], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <8 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <8 x i8> %vec, i8 %val, i32 %idx
|
|
store <8 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v16i8_s_s(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v16i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX9-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX9-NEXT: s_and_b32 s5, s5, 3
|
|
; GFX9-NEXT: s_lshl_b32 s5, s5, 3
|
|
; GFX9-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: s_movrels_b32 s6, s0
|
|
; GFX9-NEXT: s_lshl_b32 s4, s4, s5
|
|
; GFX9-NEXT: s_lshl_b32 s5, 0xff, s5
|
|
; GFX9-NEXT: s_andn2_b32 s5, s6, s5
|
|
; GFX9-NEXT: s_or_b32 s4, s5, s4
|
|
; GFX9-NEXT: s_movreld_b32 s0, s4
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v16i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX8-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX8-NEXT: s_and_b32 s5, s5, 3
|
|
; GFX8-NEXT: s_lshl_b32 s5, s5, 3
|
|
; GFX8-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_movrels_b32 s6, s0
|
|
; GFX8-NEXT: s_lshl_b32 s4, s4, s5
|
|
; GFX8-NEXT: s_lshl_b32 s5, 0xff, s5
|
|
; GFX8-NEXT: s_andn2_b32 s5, s6, s5
|
|
; GFX8-NEXT: s_or_b32 s4, s5, s4
|
|
; GFX8-NEXT: s_movreld_b32 s0, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v16i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX7-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX7-NEXT: s_and_b32 s5, s5, 3
|
|
; GFX7-NEXT: s_lshl_b32 s5, s5, 3
|
|
; GFX7-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_movrels_b32 s6, s0
|
|
; GFX7-NEXT: s_lshl_b32 s4, s4, s5
|
|
; GFX7-NEXT: s_lshl_b32 s5, 0xff, s5
|
|
; GFX7-NEXT: s_andn2_b32 s5, s6, s5
|
|
; GFX7-NEXT: s_or_b32 s4, s5, s4
|
|
; GFX7-NEXT: s_movreld_b32 s0, s4
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v16i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX10-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX10-NEXT: s_and_b32 s5, s5, 3
|
|
; GFX10-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s5, s5, 3
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX10-NEXT: s_lshl_b32 s7, 0xff, s5
|
|
; GFX10-NEXT: s_lshl_b32 s4, s4, s5
|
|
; GFX10-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: s_movrels_b32 s6, s0
|
|
; GFX10-NEXT: s_andn2_b32 s5, s6, s7
|
|
; GFX10-NEXT: s_or_b32 s4, s5, s4
|
|
; GFX10-NEXT: s_movreld_b32 s0, s4
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v16i8_s_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
|
|
; GFX11-NEXT: s_lshr_b32 m0, s5, 2
|
|
; GFX11-NEXT: s_and_b32 s5, s5, 3
|
|
; GFX11-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX11-NEXT: s_lshl_b32 s5, s5, 3
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v5, 0
|
|
; GFX11-NEXT: s_lshl_b32 s7, 0xff, s5
|
|
; GFX11-NEXT: s_lshl_b32 s4, s4, s5
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_movrels_b32 s6, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_and_not1_b32 s5, s6, s7
|
|
; GFX11-NEXT: s_or_b32 s4, s5, s4
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_movreld_b32 s0, s4
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
|
|
; GFX11-NEXT: global_store_b128 v[4:5], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v16i8_s_s(ptr addrspace(1) %ptr, i8 inreg %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v16i8_s_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s1, s3, 3
|
|
; GFX9-NEXT: s_and_b32 s2, s2, 0xff
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX9-NEXT: s_lshr_b32 s0, s3, 2
|
|
; GFX9-NEXT: s_lshl_b32 s2, s2, s1
|
|
; GFX9-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX9-NEXT: s_not_b32 s1, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, s2
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_or_b32 v6, v5, s1, v4
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v6
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v16i8_s_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx4 v[0:3], v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX8-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_movrels_b32_e32 v4, v0
|
|
; GFX8-NEXT: v_bfi_b32 v4, s0, 0, v4
|
|
; GFX8-NEXT: v_or_b32_e32 v6, s1, v4
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v16i8_s_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx4 v[0:3], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX7-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: s_lshl_b32 s1, s1, s0
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_movrels_b32_e32 v4, v0
|
|
; GFX7-NEXT: v_bfi_b32 v4, s0, 0, v4
|
|
; GFX7-NEXT: v_or_b32_e32 v4, s1, v4
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v16i8_s_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx4 v[0:3], v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX10-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX10-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX10-NEXT: s_not_b32 s1, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_movrels_b32_e32 v4, v0
|
|
; GFX10-NEXT: v_and_or_b32 v6, v4, s1, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v16i8_s_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b128 v[0:3], v[0:1], off
|
|
; GFX11-NEXT: s_and_b32 s0, s3, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s3, 2
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX11-NEXT: s_lshl_b32 s2, 0xff, s0
|
|
; GFX11-NEXT: s_lshl_b32 s0, s1, s0
|
|
; GFX11-NEXT: s_not_b32 s1, s2
|
|
; GFX11-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_movrels_b32_e32 v4, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_and_or_b32 v6, v4, s1, s0
|
|
; GFX11-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX11-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX11-NEXT: global_store_b128 v[4:5], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(1 ) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v16i8_v_s(ptr addrspace(4) inreg %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_s_v16i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX9-NEXT: s_lshr_b32 s5, s4, 2
|
|
; GFX9-NEXT: s_and_b32 s4, s4, 3
|
|
; GFX9-NEXT: s_mov_b32 m0, s5
|
|
; GFX9-NEXT: s_lshl_b32 s4, s4, 3
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: s_movrels_b32 s6, s0
|
|
; GFX9-NEXT: s_lshl_b32 s7, 0xff, s4
|
|
; GFX9-NEXT: s_andn2_b32 s6, s6, s7
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s6
|
|
; GFX9-NEXT: v_lshl_or_b32 v6, v0, s4, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s5, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, v6
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v16i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX8-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX8-NEXT: s_and_b32 s4, s4, 3
|
|
; GFX8-NEXT: s_lshl_b32 s4, s4, 3
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s4
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: s_movrels_b32 s5, s0
|
|
; GFX8-NEXT: s_lshl_b32 s4, 0xff, s4
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_andn2_b32 s4, s5, s4
|
|
; GFX8-NEXT: v_or_b32_e32 v6, s4, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v16i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX7-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX7-NEXT: s_and_b32 s4, s4, 3
|
|
; GFX7-NEXT: s_lshl_b32 s4, s4, 3
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: s_movrels_b32 s5, s0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, s4, v0
|
|
; GFX7-NEXT: s_lshl_b32 s4, 0xff, s4
|
|
; GFX7-NEXT: s_andn2_b32 s4, s5, s4
|
|
; GFX7-NEXT: v_or_b32_e32 v4, s4, v0
|
|
; GFX7-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: v_movreld_b32_e32 v0, v4
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v16i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX10-NEXT: s_and_b32 s5, s4, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX10-NEXT: s_lshl_b32 s4, s5, 3
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX10-NEXT: s_lshl_b32 s5, 0xff, s4
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: s_movrels_b32 s6, s0
|
|
; GFX10-NEXT: s_andn2_b32 s5, s6, s5
|
|
; GFX10-NEXT: v_lshl_or_b32 v6, v0, s4, s5
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX10-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v16i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
|
|
; GFX11-NEXT: s_and_b32 s5, s4, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s4, 2
|
|
; GFX11-NEXT: s_lshl_b32 s4, s5, 3
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX11-NEXT: s_lshl_b32 s5, 0xff, s4
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v5, 0
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: s_movrels_b32 s6, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: s_and_not1_b32 s5, s6, s5
|
|
; GFX11-NEXT: v_lshl_or_b32 v6, v0, s4, s5
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_movreld_b32_e32 v0, v6
|
|
; GFX11-NEXT: global_store_b128 v[4:5], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v16i8_s_v(ptr addrspace(4) inreg %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v16i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v5, 2, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v5
|
|
; GFX9-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v6, v1, v2, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v5
|
|
; GFX9-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX9-NEXT: v_mov_b32_e32 v8, 0xff
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v6, v6, v3, s[0:1]
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v5
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v7, v0, s4
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v0, v8
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v6, v6, v4, s[2:3]
|
|
; GFX9-NEXT: v_not_b32_e32 v0, v0
|
|
; GFX9-NEXT: v_and_or_b32 v7, v6, v0, v7
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v5
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v1, v7, s[4:5]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v2, v7, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v3, v7, s[0:1]
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v4, v7, s[2:3]
|
|
; GFX9-NEXT: global_store_dwordx4 v[5:6], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v16i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v5, 2, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v5
|
|
; GFX8-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v6, v1, v2, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v5
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v6, v6, v3, s[0:1]
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v5
|
|
; GFX8-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, 0xff
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v6, v6, v4, s[2:3]
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v7, v0, s4
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v0, v8
|
|
; GFX8-NEXT: v_bfi_b32 v0, v0, 0, v6
|
|
; GFX8-NEXT: v_or_b32_e32 v7, v0, v7
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v5
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v1, v7, s[4:5]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v2, v7, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v3, v7, s[0:1]
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v4, v7, s[2:3]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[5:6], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v16i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v5, 2, v0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v5
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v6, v1, v2, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v5
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v6, v6, v3, s[0:1]
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v5
|
|
; GFX7-NEXT: s_and_b32 s4, s4, 0xff
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v6, v6, v4, s[2:3]
|
|
; GFX7-NEXT: v_lshl_b32_e32 v7, s4, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_bfi_b32 v0, v0, 0, v6
|
|
; GFX7-NEXT: v_or_b32_e32 v6, v0, v7
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v5
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v1, v6, s[4:5]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v2, v6, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v2, v3, v6, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v4, v6, s[2:3]
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v16i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v6, 2, v0
|
|
; GFX10-NEXT: v_and_b32_e32 v4, 3, v0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v4, 3, v4
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v7, v4, 0xff
|
|
; GFX10-NEXT: v_not_b32_e32 v7, v7
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s2
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, s3
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v6
|
|
; GFX10-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v4, v4, s1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v6
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v6
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v2, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v5, v5, v3, s1
|
|
; GFX10-NEXT: v_and_or_b32 v7, v5, v7, v4
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, v7, s2
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v7, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v7, s1
|
|
; GFX10-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v16i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v6, 2, v0
|
|
; GFX11-NEXT: v_and_b32_e32 v4, 3, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v6
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3
|
|
; GFX11-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v4, 3, v4
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 2, v6
|
|
; GFX11-NEXT: s_and_b32 s1, s4, 0xff
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v5, v0, v1, vcc_lo
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s2, 0, v6
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v7, v4, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v4, v4, s1
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s1, 3, v6
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v5, v5, v2, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_not_b32_e32 v7, v7
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v5, v5, v3, s1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_and_or_b32 v7, v5, v7, v4
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v5, 0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, v7, s2
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v1, v7, vcc_lo
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, v7, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v3, v7, s1
|
|
; GFX11-NEXT: global_store_b128 v[4:5], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_s_v16i8_v_v(ptr addrspace(4) inreg %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_s_v16i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v6, 2, v1
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
|
|
; GFX9-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX9-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v5, s3
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX9-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX9-NEXT: v_mov_b32_e32 v4, s2
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v7, v2, v3, vcc
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v6
|
|
; GFX9-NEXT: v_mov_b32_e32 v8, 0xff
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v7, v7, v4, s[0:1]
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v6
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v1, v1, v8
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v7, v7, v5, s[2:3]
|
|
; GFX9-NEXT: v_not_b32_e32 v1, v1
|
|
; GFX9-NEXT: v_and_or_b32 v8, v7, v1, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v6
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v2, v8, s[4:5]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v8, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v4, v8, s[0:1]
|
|
; GFX9-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v5, v8, s[2:3]
|
|
; GFX9-NEXT: global_store_dwordx4 v[6:7], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_s_v16i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v6, 2, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
|
|
; GFX8-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s3
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s2
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v7, v2, v3, vcc
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v6
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v7, v7, v4, s[0:1]
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v6
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, 0xff
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v7, v7, v5, s[2:3]
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v1, v1, v8
|
|
; GFX8-NEXT: v_bfi_b32 v1, v1, 0, v7
|
|
; GFX8-NEXT: v_or_b32_e32 v8, v1, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v6
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v2, v8, s[4:5]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v3, v8, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v4, v8, s[0:1]
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v5, v8, s[2:3]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[6:7], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_s_v16i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v6, 2, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v6
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX7-NEXT: v_mov_b32_e32 v5, s3
|
|
; GFX7-NEXT: v_mov_b32_e32 v3, s1
|
|
; GFX7-NEXT: v_mov_b32_e32 v2, s0
|
|
; GFX7-NEXT: v_mov_b32_e32 v4, s2
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v7, v2, v3, vcc
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v6
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v7, v7, v4, s[0:1]
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v6
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v7, v7, v5, s[2:3]
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_bfi_b32 v1, v1, 0, v7
|
|
; GFX7-NEXT: v_or_b32_e32 v7, v1, v0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v6
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v2, v7, s[4:5]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v3, v7, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v2, v4, v7, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v5, v7, s[2:3]
|
|
; GFX7-NEXT: s_mov_b64 s[0:1], 0
|
|
; GFX7-NEXT: s_mov_b32 s2, -1
|
|
; GFX7-NEXT: s_mov_b32 s3, 0xf000
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[0:3], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_s_v16i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: s_load_dwordx4 s[0:3], s[2:3], 0x0
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v7, 2, v1
|
|
; GFX10-NEXT: v_and_b32_e32 v5, 3, v1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v7
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v5, 3, v5
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v8, v5, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_not_b32_e32 v5, v8
|
|
; GFX10-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX10-NEXT: v_mov_b32_e32 v4, s3
|
|
; GFX10-NEXT: v_mov_b32_e32 v2, s1
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, s0
|
|
; GFX10-NEXT: v_mov_b32_e32 v3, s2
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v7
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v7
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v7
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v6, v1, v2, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v3, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v6, v6, v4, s1
|
|
; GFX10-NEXT: v_and_or_b32 v8, v6, v5, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v1, v8, s2
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v2, v8, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v3, v8, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v4, v8, s1
|
|
; GFX10-NEXT: global_store_dwordx4 v[5:6], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_s_v16i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_load_b128 s[0:3], s[2:3], 0x0
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v7, 2, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v5, 3, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v7
|
|
; GFX11-NEXT: s_waitcnt lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, s3 :: v_dual_mov_b32 v3, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s1 :: v_dual_mov_b32 v1, s0
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v5, 3, v5
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 2, v7
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s1, 3, v7
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s2, 0, v7
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v6, v1, v2, vcc_lo
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v8, v5, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, v5, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v6, v6, v3, s0
|
|
; GFX11-NEXT: v_not_b32_e32 v5, v8
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v6, v6, v4, s1
|
|
; GFX11-NEXT: v_and_or_b32 v8, v6, v5, v0
|
|
; GFX11-NEXT: v_dual_mov_b32 v5, 0 :: v_dual_mov_b32 v6, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v1, v8, s2
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v2, v8, vcc_lo
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v3, v8, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v4, v8, s1
|
|
; GFX11-NEXT: global_store_b128 v[5:6], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(4) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v16i8_s_v(ptr addrspace(1) %ptr, i8 inreg %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v16i8_s_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 2, v2
|
|
; GFX9-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0xff
|
|
; GFX9-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX9-NEXT: v_lshlrev_b32_e64 v9, v2, s0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v2, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v1
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v1
|
|
; GFX9-NEXT: v_not_b32_e32 v0, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v2, v5, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v2, v6, s[2:3]
|
|
; GFX9-NEXT: v_and_or_b32 v9, v2, v0, v9
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v3, v9, s[4:5]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v4, v9, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v5, v9, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v6, v9, s[2:3]
|
|
; GFX9-NEXT: global_store_dwordx4 v[7:8], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v16i8_s_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx4 v[3:6], v[0:1]
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 2, v2
|
|
; GFX8-NEXT: v_and_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0xff
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v2, 3, v2
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_e64 v9, v2, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v2, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v2, v5, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v2, v6, s[2:3]
|
|
; GFX8-NEXT: v_bfi_b32 v0, v0, 0, v2
|
|
; GFX8-NEXT: v_or_b32_e32 v9, v0, v9
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v3, v9, s[4:5]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v4, v9, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v5, v9, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v6, v9, s[2:3]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[7:8], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v16i8_s_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s10, 0
|
|
; GFX7-NEXT: s_mov_b32 s11, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[8:9], 0
|
|
; GFX7-NEXT: buffer_load_dwordx4 v[3:6], v[0:1], s[8:11], 0 addr64
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 2, v2
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v2
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 0xff
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v2, s0, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v0
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v0
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v0
|
|
; GFX7-NEXT: s_mov_b64 s[8:9], 0
|
|
; GFX7-NEXT: s_mov_b32 s10, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v7, v3, v4, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v7, v7, v5, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v7, v7, v6, s[2:3]
|
|
; GFX7-NEXT: v_bfi_b32 v1, v1, 0, v7
|
|
; GFX7-NEXT: v_or_b32_e32 v7, v1, v2
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v3, v7, s[4:5]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v4, v7, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v2, v5, v7, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v6, v7, s[2:3]
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v16i8_s_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 2, v2
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 3, v2
|
|
; GFX10-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX10-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v7, v0, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v0, v0, s1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
|
|
; GFX10-NEXT: v_not_b32_e32 v7, v7
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v2, v6, s1
|
|
; GFX10-NEXT: v_and_or_b32 v9, v2, v7, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v3, v9, s2
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v4, v9, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v5, v9, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v6, v9, s1
|
|
; GFX10-NEXT: global_store_dwordx4 v[7:8], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v16i8_s_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b128 v[3:6], v[0:1], off
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v1, 2, v2
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 3, v2
|
|
; GFX11-NEXT: s_and_b32 s1, s2, 0xff
|
|
; GFX11-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_2) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v7, v0, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v0, v0, s1
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
|
|
; GFX11-NEXT: v_not_b32_e32 v7, v7
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v2, v3, v4, vcc_lo
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, v5, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v2, v6, s1
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_and_or_b32 v9, v2, v7, v0
|
|
; GFX11-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v3, v9, s2
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v4, v9, vcc_lo
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v5, v9, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v6, v9, s1
|
|
; GFX11-NEXT: global_store_b128 v[7:8], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v16i8_v_s(ptr addrspace(1) %ptr, i8 %val, i32 inreg %idx) {
|
|
; GFX9-LABEL: insertelement_v_v16i8_v_s:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
|
|
; GFX9-NEXT: s_and_b32 s1, s2, 3
|
|
; GFX9-NEXT: s_lshl_b32 s1, s1, 3
|
|
; GFX9-NEXT: s_lshr_b32 s0, s2, 2
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v0, s1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: s_lshl_b32 s1, 0xff, s1
|
|
; GFX9-NEXT: s_not_b32 s1, s1
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(SRC0)
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: v_and_or_b32 v2, v1, s1, v0
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX9-NEXT: s_set_gpr_idx_on s0, gpr_idx(DST)
|
|
; GFX9-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX9-NEXT: s_set_gpr_idx_off
|
|
; GFX9-NEXT: global_store_dwordx4 v[0:1], v[3:6], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v16i8_v_s:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx4 v[3:6], v[0:1]
|
|
; GFX8-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX8-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX8-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s0
|
|
; GFX8-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_movrels_b32_e32 v1, v3
|
|
; GFX8-NEXT: v_bfi_b32 v1, s0, 0, v1
|
|
; GFX8-NEXT: v_or_b32_e32 v2, v1, v0
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX8-NEXT: v_movreld_b32_e32 v3, v2
|
|
; GFX8-NEXT: flat_store_dwordx4 v[0:1], v[3:6]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v16i8_v_s:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s6, 0
|
|
; GFX7-NEXT: s_mov_b32 s7, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: buffer_load_dwordx4 v[3:6], v[0:1], s[4:7], 0 addr64
|
|
; GFX7-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX7-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX7-NEXT: v_and_b32_e32 v0, 0xff, v2
|
|
; GFX7-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v0, s0, v0
|
|
; GFX7-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX7-NEXT: s_mov_b64 s[4:5], 0
|
|
; GFX7-NEXT: s_mov_b32 s6, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_movrels_b32_e32 v1, v3
|
|
; GFX7-NEXT: v_bfi_b32 v1, s0, 0, v1
|
|
; GFX7-NEXT: v_or_b32_e32 v0, v1, v0
|
|
; GFX7-NEXT: v_movreld_b32_e32 v3, v0
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[3:6], off, s[4:7], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v16i8_v_s:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx4 v[3:6], v[0:1], off
|
|
; GFX10-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX10-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX10-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v1, s0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX10-NEXT: s_not_b32 s0, s0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_movrels_b32_e32 v0, v3
|
|
; GFX10-NEXT: v_and_or_b32 v2, v0, s0, v1
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX10-NEXT: v_movreld_b32_e32 v3, v2
|
|
; GFX10-NEXT: global_store_dwordx4 v[0:1], v[3:6], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v16i8_v_s:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b128 v[3:6], v[0:1], off
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 0xff, v2
|
|
; GFX11-NEXT: s_and_b32 s0, s2, 3
|
|
; GFX11-NEXT: s_lshr_b32 m0, s2, 2
|
|
; GFX11-NEXT: s_lshl_b32 s0, s0, 3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, s0, v0
|
|
; GFX11-NEXT: s_lshl_b32 s0, 0xff, s0
|
|
; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: s_not_b32 s0, s0
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_movrels_b32_e32 v1, v3
|
|
; GFX11-NEXT: v_and_or_b32 v2, v1, s0, v0
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_movreld_b32_e32 v3, v2
|
|
; GFX11-NEXT: global_store_b128 v[0:1], v[3:6], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @insertelement_v_v16i8_v_v(ptr addrspace(1) %ptr, i8 %val, i32 %idx) {
|
|
; GFX9-LABEL: insertelement_v_v16i8_v_v:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: global_load_dwordx4 v[4:7], v[0:1], off
|
|
; GFX9-NEXT: v_lshrrev_b32_e32 v1, 2, v3
|
|
; GFX9-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX9-NEXT: v_mov_b32_e32 v0, 0xff
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX9-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX9-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX9-NEXT: v_lshlrev_b32_e32 v0, v3, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v1
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v1
|
|
; GFX9-NEXT: v_not_b32_e32 v0, v0
|
|
; GFX9-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v1
|
|
; GFX9-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX9-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v3, v7, s[2:3]
|
|
; GFX9-NEXT: v_and_or_b32 v3, v3, v0, v2
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v0, v4, v3, s[4:5]
|
|
; GFX9-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v2, v6, v3, s[0:1]
|
|
; GFX9-NEXT: v_cndmask_b32_e64 v3, v7, v3, s[2:3]
|
|
; GFX9-NEXT: global_store_dwordx4 v[8:9], v[0:3], off
|
|
; GFX9-NEXT: s_endpgm
|
|
;
|
|
; GFX8-LABEL: insertelement_v_v16i8_v_v:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: flat_load_dwordx4 v[4:7], v[0:1]
|
|
; GFX8-NEXT: v_lshrrev_b32_e32 v1, 2, v3
|
|
; GFX8-NEXT: v_and_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, 0xff
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v3, 3, v3
|
|
; GFX8-NEXT: v_cmp_eq_u32_e32 vcc, 1, v1
|
|
; GFX8-NEXT: v_lshlrev_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX8-NEXT: v_lshlrev_b32_e32 v0, v3, v0
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v1
|
|
; GFX8-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX8-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v3, v7, s[2:3]
|
|
; GFX8-NEXT: v_bfi_b32 v0, v0, 0, v3
|
|
; GFX8-NEXT: v_or_b32_e32 v3, v0, v2
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v0, v4, v3, s[4:5]
|
|
; GFX8-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v2, v6, v3, s[0:1]
|
|
; GFX8-NEXT: v_cndmask_b32_e64 v3, v7, v3, s[2:3]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[8:9], v[0:3]
|
|
; GFX8-NEXT: s_endpgm
|
|
;
|
|
; GFX7-LABEL: insertelement_v_v16i8_v_v:
|
|
; GFX7: ; %bb.0:
|
|
; GFX7-NEXT: s_mov_b32 s10, 0
|
|
; GFX7-NEXT: s_mov_b32 s11, 0xf000
|
|
; GFX7-NEXT: s_mov_b64 s[8:9], 0
|
|
; GFX7-NEXT: buffer_load_dwordx4 v[4:7], v[0:1], s[8:11], 0 addr64
|
|
; GFX7-NEXT: v_lshrrev_b32_e32 v0, 2, v3
|
|
; GFX7-NEXT: v_cmp_eq_u32_e32 vcc, 1, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v1, 3, v3
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[0:1], 2, v0
|
|
; GFX7-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v1, 3, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[2:3], 3, v0
|
|
; GFX7-NEXT: v_lshlrev_b32_e32 v2, v1, v2
|
|
; GFX7-NEXT: v_lshl_b32_e32 v1, 0xff, v1
|
|
; GFX7-NEXT: v_cmp_eq_u32_e64 s[4:5], 0, v0
|
|
; GFX7-NEXT: s_mov_b64 s[8:9], 0
|
|
; GFX7-NEXT: s_mov_b32 s10, -1
|
|
; GFX7-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v3, v6, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v3, v7, s[2:3]
|
|
; GFX7-NEXT: v_bfi_b32 v1, v1, 0, v3
|
|
; GFX7-NEXT: v_or_b32_e32 v3, v1, v2
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v0, v4, v3, s[4:5]
|
|
; GFX7-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v2, v6, v3, s[0:1]
|
|
; GFX7-NEXT: v_cndmask_b32_e64 v3, v7, v3, s[2:3]
|
|
; GFX7-NEXT: buffer_store_dwordx4 v[0:3], off, s[8:11], 0
|
|
; GFX7-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: insertelement_v_v16i8_v_v:
|
|
; GFX10: ; %bb.0:
|
|
; GFX10-NEXT: global_load_dwordx4 v[4:7], v[0:1], off
|
|
; GFX10-NEXT: v_lshrrev_b32_e32 v1, 2, v3
|
|
; GFX10-NEXT: v_and_b32_e32 v0, 3, v3
|
|
; GFX10-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e32 v0, 3, v0
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
|
|
; GFX10-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
|
|
; GFX10-NEXT: v_lshlrev_b32_e64 v8, v0, 0xff
|
|
; GFX10-NEXT: v_lshlrev_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
|
|
; GFX10-NEXT: v_not_b32_e32 v2, v8
|
|
; GFX10-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX10-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v3, v4, v5, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v6, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v3, v7, s1
|
|
; GFX10-NEXT: v_and_or_b32 v3, v3, v2, v0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v0, v4, v3, s2
|
|
; GFX10-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc_lo
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v2, v6, v3, s0
|
|
; GFX10-NEXT: v_cndmask_b32_e64 v3, v7, v3, s1
|
|
; GFX10-NEXT: global_store_dwordx4 v[8:9], v[0:3], off
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX11-LABEL: insertelement_v_v16i8_v_v:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: global_load_b128 v[4:7], v[0:1], off
|
|
; GFX11-NEXT: v_lshrrev_b32_e32 v1, 2, v3
|
|
; GFX11-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v0, 3, v3
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s0, 2, v1
|
|
; GFX11-NEXT: v_and_b32_e32 v2, 0xff, v2
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s1, 3, v1
|
|
; GFX11-NEXT: v_cmp_eq_u32_e64 s2, 0, v1
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX11-NEXT: v_dual_cndmask_b32 v3, v4, v5 :: v_dual_lshlrev_b32 v0, 3, v0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v3, v6, s0
|
|
; GFX11-NEXT: v_lshlrev_b32_e64 v8, v0, 0xff
|
|
; GFX11-NEXT: v_lshlrev_b32_e32 v0, v0, v2
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v3, v7, s1
|
|
; GFX11-NEXT: v_not_b32_e32 v2, v8
|
|
; GFX11-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
|
|
; GFX11-NEXT: v_and_or_b32 v3, v3, v2, v0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v0, v4, v3, s2
|
|
; GFX11-NEXT: v_cndmask_b32_e32 v1, v5, v3, vcc_lo
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v2, v6, v3, s0
|
|
; GFX11-NEXT: v_cndmask_b32_e64 v3, v7, v3, s1
|
|
; GFX11-NEXT: global_store_b128 v[8:9], v[0:3], off
|
|
; GFX11-NEXT: s_endpgm
|
|
%vec = load <16 x i8>, ptr addrspace(1) %ptr
|
|
%insert = insertelement <16 x i8> %vec, i8 %val, i32 %idx
|
|
store <16 x i8> %insert, ptr addrspace(1) null
|
|
ret void
|
|
}
|