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
236 lines
9.1 KiB
LLVM
236 lines
9.1 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck %s --check-prefix=GFX900
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck %s --check-prefix=GFX90A
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck %s --check-prefix=GFX942
|
|
; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck %s --check-prefix=GFX10
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck %s --check-prefix=GFX900-GISEL
|
|
|
|
declare void @llvm.amdgcn.global.load.lds(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr, i32 %size, i32 %offset, i32 %aux)
|
|
|
|
define amdgpu_ps void @global_load_lds_dword_vaddr(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr) {
|
|
; GFX900-LABEL: global_load_lds_dword_vaddr:
|
|
; GFX900: ; %bb.0: ; %main_body
|
|
; GFX900-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: global_load_dword v[0:1], off offset:16 glc lds
|
|
; GFX900-NEXT: s_endpgm
|
|
;
|
|
; GFX90A-LABEL: global_load_lds_dword_vaddr:
|
|
; GFX90A: ; %bb.0: ; %main_body
|
|
; GFX90A-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX90A-NEXT: s_mov_b32 m0, s0
|
|
; GFX90A-NEXT: s_nop 0
|
|
; GFX90A-NEXT: global_load_dword v[0:1], off offset:16 glc lds
|
|
; GFX90A-NEXT: s_endpgm
|
|
;
|
|
; GFX942-LABEL: global_load_lds_dword_vaddr:
|
|
; GFX942: ; %bb.0: ; %main_body
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: s_mov_b32 m0, s0
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: global_load_lds_dword v[0:1], off offset:16 sc0
|
|
; GFX942-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: global_load_lds_dword_vaddr:
|
|
; GFX10: ; %bb.0: ; %main_body
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX10-NEXT: s_mov_b32 m0, s0
|
|
; GFX10-NEXT: global_load_dword v[0:1], off offset:16 glc lds
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX900-GISEL-LABEL: global_load_lds_dword_vaddr:
|
|
; GFX900-GISEL: ; %bb.0: ; %main_body
|
|
; GFX900-GISEL-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-GISEL-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-GISEL-NEXT: s_nop 0
|
|
; GFX900-GISEL-NEXT: global_load_dword v[0:1], off offset:16 glc lds
|
|
; GFX900-GISEL-NEXT: s_endpgm
|
|
main_body:
|
|
call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 4, i32 16, i32 1)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @global_load_lds_dword_saddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr) {
|
|
; GFX900-LABEL: global_load_lds_dword_saddr:
|
|
; GFX900: ; %bb.0: ; %main_body
|
|
; GFX900-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX900-NEXT: s_mov_b32 m0, s2
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: global_load_dword v1, s[0:1] offset:32 slc lds
|
|
; GFX900-NEXT: s_endpgm
|
|
;
|
|
; GFX90A-LABEL: global_load_lds_dword_saddr:
|
|
; GFX90A: ; %bb.0: ; %main_body
|
|
; GFX90A-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX90A-NEXT: s_mov_b32 m0, s2
|
|
; GFX90A-NEXT: s_nop 0
|
|
; GFX90A-NEXT: global_load_dword v1, s[0:1] offset:32 slc lds
|
|
; GFX90A-NEXT: s_endpgm
|
|
;
|
|
; GFX942-LABEL: global_load_lds_dword_saddr:
|
|
; GFX942: ; %bb.0: ; %main_body
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX942-NEXT: s_mov_b32 m0, s2
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: global_load_lds_dword v1, s[0:1] offset:32 nt
|
|
; GFX942-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: global_load_lds_dword_saddr:
|
|
; GFX10: ; %bb.0: ; %main_body
|
|
; GFX10-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX10-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX10-NEXT: s_mov_b32 m0, s2
|
|
; GFX10-NEXT: global_load_dword v0, s[0:1] offset:32 slc lds
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX900-GISEL-LABEL: global_load_lds_dword_saddr:
|
|
; GFX900-GISEL: ; %bb.0: ; %main_body
|
|
; GFX900-GISEL-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX900-GISEL-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-GISEL-NEXT: s_mov_b32 m0, s2
|
|
; GFX900-GISEL-NEXT: s_nop 0
|
|
; GFX900-GISEL-NEXT: global_load_dword v0, s[0:1] offset:32 slc lds
|
|
; GFX900-GISEL-NEXT: s_endpgm
|
|
main_body:
|
|
call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 4, i32 32, i32 2)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @global_load_lds_dword_saddr_and_vaddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr, i32 %voffset) {
|
|
; GFX900-LABEL: global_load_lds_dword_saddr_and_vaddr:
|
|
; GFX900: ; %bb.0: ; %main_body
|
|
; GFX900-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX900-NEXT: s_mov_b32 m0, s2
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: global_load_dword v1, s[0:1] offset:48 lds
|
|
; GFX900-NEXT: s_endpgm
|
|
;
|
|
; GFX90A-LABEL: global_load_lds_dword_saddr_and_vaddr:
|
|
; GFX90A: ; %bb.0: ; %main_body
|
|
; GFX90A-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX90A-NEXT: s_mov_b32 m0, s2
|
|
; GFX90A-NEXT: s_nop 0
|
|
; GFX90A-NEXT: global_load_dword v1, s[0:1] offset:48 scc lds
|
|
; GFX90A-NEXT: s_endpgm
|
|
;
|
|
; GFX942-LABEL: global_load_lds_dword_saddr_and_vaddr:
|
|
; GFX942: ; %bb.0: ; %main_body
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX942-NEXT: s_mov_b32 m0, s2
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: global_load_lds_dword v1, s[0:1] offset:48 sc1
|
|
; GFX942-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: global_load_lds_dword_saddr_and_vaddr:
|
|
; GFX10: ; %bb.0: ; %main_body
|
|
; GFX10-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX10-NEXT: s_mov_b32 m0, s2
|
|
; GFX10-NEXT: global_load_dword v1, s[0:1] offset:48 lds
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX900-GISEL-LABEL: global_load_lds_dword_saddr_and_vaddr:
|
|
; GFX900-GISEL: ; %bb.0: ; %main_body
|
|
; GFX900-GISEL-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX900-GISEL-NEXT: s_mov_b32 m0, s2
|
|
; GFX900-GISEL-NEXT: s_nop 0
|
|
; GFX900-GISEL-NEXT: global_load_dword v1, s[0:1] offset:48 lds
|
|
; GFX900-GISEL-NEXT: s_endpgm
|
|
main_body:
|
|
%voffset.64 = zext i32 %voffset to i64
|
|
%gep = getelementptr i8, ptr addrspace(1) %gptr, i64 %voffset.64
|
|
call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gep, ptr addrspace(3) %lptr, i32 4, i32 48, i32 16)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @global_load_lds_ushort_vaddr(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr) {
|
|
; GFX900-LABEL: global_load_lds_ushort_vaddr:
|
|
; GFX900: ; %bb.0: ; %main_body
|
|
; GFX900-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: global_load_ushort v[0:1], off lds
|
|
; GFX900-NEXT: s_endpgm
|
|
;
|
|
; GFX90A-LABEL: global_load_lds_ushort_vaddr:
|
|
; GFX90A: ; %bb.0: ; %main_body
|
|
; GFX90A-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX90A-NEXT: s_mov_b32 m0, s0
|
|
; GFX90A-NEXT: s_nop 0
|
|
; GFX90A-NEXT: global_load_ushort v[0:1], off lds
|
|
; GFX90A-NEXT: s_endpgm
|
|
;
|
|
; GFX942-LABEL: global_load_lds_ushort_vaddr:
|
|
; GFX942: ; %bb.0: ; %main_body
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: s_mov_b32 m0, s0
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: global_load_lds_ushort v[0:1], off
|
|
; GFX942-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: global_load_lds_ushort_vaddr:
|
|
; GFX10: ; %bb.0: ; %main_body
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX10-NEXT: s_mov_b32 m0, s0
|
|
; GFX10-NEXT: global_load_ushort v[0:1], off dlc lds
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX900-GISEL-LABEL: global_load_lds_ushort_vaddr:
|
|
; GFX900-GISEL: ; %bb.0: ; %main_body
|
|
; GFX900-GISEL-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-GISEL-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-GISEL-NEXT: s_nop 0
|
|
; GFX900-GISEL-NEXT: global_load_ushort v[0:1], off lds
|
|
; GFX900-GISEL-NEXT: s_endpgm
|
|
main_body:
|
|
call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 2, i32 0, i32 4)
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_ps void @global_load_lds_ubyte_vaddr(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr) {
|
|
; GFX900-LABEL: global_load_lds_ubyte_vaddr:
|
|
; GFX900: ; %bb.0: ; %main_body
|
|
; GFX900-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: global_load_ubyte v[0:1], off lds
|
|
; GFX900-NEXT: s_endpgm
|
|
;
|
|
; GFX90A-LABEL: global_load_lds_ubyte_vaddr:
|
|
; GFX90A: ; %bb.0: ; %main_body
|
|
; GFX90A-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX90A-NEXT: s_mov_b32 m0, s0
|
|
; GFX90A-NEXT: s_nop 0
|
|
; GFX90A-NEXT: global_load_ubyte v[0:1], off lds
|
|
; GFX90A-NEXT: s_endpgm
|
|
;
|
|
; GFX942-LABEL: global_load_lds_ubyte_vaddr:
|
|
; GFX942: ; %bb.0: ; %main_body
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: s_mov_b32 m0, s0
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: global_load_lds_ubyte v[0:1], off
|
|
; GFX942-NEXT: s_endpgm
|
|
;
|
|
; GFX10-LABEL: global_load_lds_ubyte_vaddr:
|
|
; GFX10: ; %bb.0: ; %main_body
|
|
; GFX10-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX10-NEXT: s_mov_b32 m0, s0
|
|
; GFX10-NEXT: global_load_ubyte v[0:1], off lds
|
|
; GFX10-NEXT: s_endpgm
|
|
;
|
|
; GFX900-GISEL-LABEL: global_load_lds_ubyte_vaddr:
|
|
; GFX900-GISEL: ; %bb.0: ; %main_body
|
|
; GFX900-GISEL-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX900-GISEL-NEXT: s_mov_b32 m0, s0
|
|
; GFX900-GISEL-NEXT: s_nop 0
|
|
; GFX900-GISEL-NEXT: global_load_ubyte v[0:1], off lds
|
|
; GFX900-GISEL-NEXT: s_endpgm
|
|
main_body:
|
|
call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 1, i32 0, i32 0)
|
|
ret void
|
|
}
|