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
146 lines
10 KiB
LLVM
146 lines
10 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -O0 -mtriple=amdgcn -mcpu=gfx1100 -start-before=amdgpu-isel -stop-after=amdgpu-isel < %s | FileCheck --check-prefixes=CHECK-O0 %s
|
|
; RUN: llc -O1 -mtriple=amdgcn -mcpu=gfx1100 -start-before=amdgpu-isel -stop-after=amdgpu-isel < %s | FileCheck --check-prefixes=CHECK-O1 %s
|
|
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128:128:48-p9:192:256:256:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7:8:9"
|
|
target triple = "amdgcn-amd-amdhsa"
|
|
|
|
define void @_QFPadd(ptr %0, ptr %1) #0 !dbg !12 {
|
|
; CHECK-O0-LABEL: name: _QFPadd
|
|
; CHECK-O0: bb.0 (%ir-block.2):
|
|
; CHECK-O0-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)
|
|
; CHECK-O0-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
|
|
; CHECK-O0-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
|
|
; CHECK-O0-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
|
|
; CHECK-O0-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; CHECK-O0-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1
|
|
; CHECK-O0-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1
|
|
; CHECK-O0-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]]
|
|
; CHECK-O0-NEXT: DBG_VALUE [[COPY4]], 0, !9, !DIExpression(DIOpArg(0, ptr), DIOpDeref(ptr)), debug-location !8
|
|
; CHECK-O0-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
|
|
; CHECK-O0-NEXT: DBG_VALUE [[COPY5]], 0, !7, !DIExpression(DIOpArg(0, ptr), DIOpDeref(ptr)), debug-location !8
|
|
; CHECK-O0-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]], debug-location !10
|
|
; CHECK-O0-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY6]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !10 :: (load (s32) from %ir.0)
|
|
; CHECK-O0-NEXT: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]], debug-location !10
|
|
; CHECK-O0-NEXT: [[FLAT_LOAD_DWORD1:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY7]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !10 :: (load (s32) from %ir.1)
|
|
; CHECK-O0-NEXT: [[V_CMP_GT_I32_e64_:%[0-9]+]]:sreg_32 = V_CMP_GT_I32_e64 killed [[FLAT_LOAD_DWORD]], killed [[FLAT_LOAD_DWORD1]], implicit $exec, debug-location !10
|
|
; CHECK-O0-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 -1
|
|
; CHECK-O0-NEXT: [[S_XOR_B32_:%[0-9]+]]:sreg_32 = S_XOR_B32 killed [[V_CMP_GT_I32_e64_]], killed [[S_MOV_B32_]], implicit-def dead $scc, debug-location !10
|
|
; CHECK-O0-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 $exec_lo, killed [[S_XOR_B32_]], implicit-def dead $scc, debug-location !10
|
|
; CHECK-O0-NEXT: $vcc_lo = COPY [[S_AND_B32_]], debug-location !10
|
|
; CHECK-O0-NEXT: S_CBRANCH_VCCNZ %bb.2, implicit $vcc, debug-location !10
|
|
; CHECK-O0-NEXT: S_BRANCH %bb.1, debug-location !10
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: bb.1 (%ir-block.6):
|
|
; CHECK-O0-NEXT: successors: %bb.3(0x80000000)
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: [[FLAT_LOAD_DWORD2:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY5]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !11 :: (load (s32) from %ir.0)
|
|
; CHECK-O0-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 1
|
|
; CHECK-O0-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[FLAT_LOAD_DWORD2]], killed [[S_MOV_B32_1]], 0, implicit $exec, debug-location !11
|
|
; CHECK-O0-NEXT: FLAT_STORE_DWORD [[COPY4]], killed [[V_ADD_U32_e64_]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !11 :: (store (s32) into %ir.1)
|
|
; CHECK-O0-NEXT: S_BRANCH %bb.3, debug-location !10
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: bb.2 (%ir-block.9):
|
|
; CHECK-O0-NEXT: successors: %bb.3(0x80000000)
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: [[FLAT_LOAD_DWORD3:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY4]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !12 :: (load (s32) from %ir.1)
|
|
; CHECK-O0-NEXT: [[S_MOV_B32_2:%[0-9]+]]:sreg_32 = S_MOV_B32 1
|
|
; CHECK-O0-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[FLAT_LOAD_DWORD3]], killed [[S_MOV_B32_2]], 0, implicit $exec, debug-location !12
|
|
; CHECK-O0-NEXT: FLAT_STORE_DWORD [[COPY4]], killed [[V_ADD_U32_e64_1]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !12 :: (store (s32) into %ir.1)
|
|
; CHECK-O0-NEXT: S_BRANCH %bb.3, debug-location !10
|
|
; CHECK-O0-NEXT: {{ $}}
|
|
; CHECK-O0-NEXT: bb.3 (%ir-block.12):
|
|
; CHECK-O0-NEXT: SI_RETURN debug-location !13
|
|
;
|
|
; CHECK-O1-LABEL: name: _QFPadd
|
|
; CHECK-O1: bb.0 (%ir-block.2):
|
|
; CHECK-O1-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000)
|
|
; CHECK-O1-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: DBG_PHI $vgpr1, 6
|
|
; CHECK-O1-NEXT: DBG_PHI $vgpr0, 5
|
|
; CHECK-O1-NEXT: DBG_PHI $vgpr3, 3
|
|
; CHECK-O1-NEXT: DBG_PHI $vgpr2, 2
|
|
; CHECK-O1-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr3
|
|
; CHECK-O1-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr2
|
|
; CHECK-O1-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr1
|
|
; CHECK-O1-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr0
|
|
; CHECK-O1-NEXT: [[REG_SEQUENCE:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, [[COPY]], %subreg.sub1, debug-instr-number 1
|
|
; CHECK-O1-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sreg_64 = REG_SEQUENCE [[COPY3]], %subreg.sub0, [[COPY2]], %subreg.sub1, debug-instr-number 4
|
|
; CHECK-O1-NEXT: [[COPY4:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]]
|
|
; CHECK-O1-NEXT: DBG_INSTR_REF !9, !DIExpression(DIOpArg(0, i32), DIOpArg(1, i32), DIOpComposite(2, ptr), DIOpDeref(ptr)), dbg-instr-ref(2, 0), dbg-instr-ref(3, 0), debug-location !8
|
|
; CHECK-O1-NEXT: [[COPY5:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]]
|
|
; CHECK-O1-NEXT: DBG_INSTR_REF !7, !DIExpression(DIOpArg(0, i32), DIOpArg(1, i32), DIOpComposite(2, ptr), DIOpDeref(ptr)), dbg-instr-ref(5, 0), dbg-instr-ref(6, 0), debug-location !8
|
|
; CHECK-O1-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE1]], debug-location !10
|
|
; CHECK-O1-NEXT: [[FLAT_LOAD_DWORD:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY6]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !10 :: (load (s32) from %ir.0)
|
|
; CHECK-O1-NEXT: [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]], debug-location !10
|
|
; CHECK-O1-NEXT: [[FLAT_LOAD_DWORD1:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY7]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !10 :: (load (s32) from %ir.1)
|
|
; CHECK-O1-NEXT: [[V_CMP_LE_I32_e64_:%[0-9]+]]:sreg_32 = V_CMP_LE_I32_e64 killed [[FLAT_LOAD_DWORD]], killed [[FLAT_LOAD_DWORD1]], implicit $exec, debug-location !10
|
|
; CHECK-O1-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 $exec_lo, killed [[V_CMP_LE_I32_e64_]], implicit-def dead $scc, debug-location !10
|
|
; CHECK-O1-NEXT: $vcc_lo = COPY [[S_AND_B32_]], debug-location !10
|
|
; CHECK-O1-NEXT: S_CBRANCH_VCCNZ %bb.2, implicit $vcc, debug-location !10
|
|
; CHECK-O1-NEXT: S_BRANCH %bb.1, debug-location !10
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: bb.1 (%ir-block.6):
|
|
; CHECK-O1-NEXT: successors: %bb.3(0x80000000)
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: [[FLAT_LOAD_DWORD2:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY5]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !11 :: (load (s32) from %ir.0)
|
|
; CHECK-O1-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 1
|
|
; CHECK-O1-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[FLAT_LOAD_DWORD2]], killed [[S_MOV_B32_]], 0, implicit $exec, debug-location !11
|
|
; CHECK-O1-NEXT: FLAT_STORE_DWORD [[COPY4]], killed [[V_ADD_U32_e64_]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !11 :: (store (s32) into %ir.1)
|
|
; CHECK-O1-NEXT: S_BRANCH %bb.3, debug-location !10
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: bb.2 (%ir-block.9):
|
|
; CHECK-O1-NEXT: successors: %bb.3(0x80000000)
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: [[FLAT_LOAD_DWORD3:%[0-9]+]]:vgpr_32 = FLAT_LOAD_DWORD [[COPY4]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !12 :: (load (s32) from %ir.1)
|
|
; CHECK-O1-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sreg_32 = S_MOV_B32 1
|
|
; CHECK-O1-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 killed [[FLAT_LOAD_DWORD3]], killed [[S_MOV_B32_1]], 0, implicit $exec, debug-location !12
|
|
; CHECK-O1-NEXT: FLAT_STORE_DWORD [[COPY4]], killed [[V_ADD_U32_e64_1]], 0, 0, implicit $exec, implicit $flat_scr, debug-location !12 :: (store (s32) into %ir.1)
|
|
; CHECK-O1-NEXT: {{ $}}
|
|
; CHECK-O1-NEXT: bb.3 (%ir-block.12):
|
|
; CHECK-O1-NEXT: SI_RETURN debug-location !13
|
|
#dbg_declare(ptr %0, !17, !DIExpression(DIOpArg(0, ptr), DIOpDeref(ptr)), !18)
|
|
#dbg_declare(ptr %1, !19, !DIExpression(DIOpArg(0, ptr), DIOpDeref(ptr)), !18)
|
|
%3 = load i32, ptr %0, align 4, !dbg !20
|
|
%4 = load i32, ptr %1, align 4, !dbg !20
|
|
%5 = icmp sgt i32 %3, %4, !dbg !20
|
|
br i1 %5, label %6, label %9, !dbg !20
|
|
|
|
6: ; preds = %2
|
|
%7 = load i32, ptr %0, align 4, !dbg !21
|
|
%8 = add i32 %7, 1, !dbg !21
|
|
store i32 %8, ptr %1, align 4, !dbg !21
|
|
br label %12, !dbg !20
|
|
|
|
9: ; preds = %2
|
|
%10 = load i32, ptr %1, align 4, !dbg !22
|
|
%11 = add i32 %10, 1, !dbg !22
|
|
store i32 %11, ptr %1, align 4, !dbg !22
|
|
br label %12, !dbg !20
|
|
|
|
12: ; preds = %9, %6
|
|
ret void, !dbg !23
|
|
}
|
|
|
|
|
|
!llvm.module.flags = !{!2}
|
|
!llvm.dbg.cu = !{!6}
|
|
|
|
!2 = !{i32 2, !"Debug Info Version", i32 3}
|
|
!6 = distinct !DICompileUnit(language: DW_LANG_Fortran95, file: !7, producer: "flang", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug)
|
|
!7 = !DIFile(filename: "target14.f90", directory: "")
|
|
!11 = !{i32 2, i32 0}
|
|
!12 = distinct !DISubprogram(name: "add", linkageName: "_QFPadd", scope: !7, file: !7, line: 16, type: !14, scopeLine: 16, spFlags: DISPFlagDefinition, unit: !6)
|
|
!14 = !DISubroutineType(cc: DW_CC_normal, types: !15)
|
|
!15 = !{null, !16, !16}
|
|
!16 = !DIBasicType(name: "integer", size: 32, encoding: DW_ATE_signed)
|
|
!17 = !DILocalVariable(name: "a", arg: 1, scope: !12, file: !7, line: 17, type: !16)
|
|
!18 = !DILocation(line: 16, column: 7, scope: !12)
|
|
!19 = !DILocalVariable(name: "b", arg: 2, scope: !12, file: !7, line: 17, type: !16)
|
|
!20 = !DILocation(line: 20, column: 7, scope: !12)
|
|
!21 = !DILocation(line: 21, column: 7, scope: !12)
|
|
!22 = !DILocation(line: 23, column: 7, scope: !12)
|
|
!23 = !DILocation(line: 25, column: 7, scope: !12)
|