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
206 lines
9.7 KiB
LLVM
206 lines
9.7 KiB
LLVM
; RUN: opt -S -passes=loop-vectorize -pass-remarks-missed=loop-vectorize \
|
|
; RUN: -pass-remarks-with-hotness < %s 2>&1 | \
|
|
; RUN: FileCheck -check-prefix=HOTNESS -check-prefix=BOTH %s
|
|
|
|
; RUN: opt -S -passes=loop-vectorize -pass-remarks-missed=loop-vectorize < %s 2>&1 | \
|
|
; RUN: FileCheck -check-prefix=NO_HOTNESS -check-prefix=BOTH %s
|
|
|
|
|
|
; RUN: opt -S -passes=loop-vectorize -pass-remarks-missed=loop-vectorize \
|
|
; RUN: -pass-remarks-with-hotness < %s 2>&1 | \
|
|
; RUN: FileCheck -check-prefix=HOTNESS -check-prefix=BOTH %s
|
|
|
|
; RUN: opt -S -passes=loop-vectorize \
|
|
; RUN: -pass-remarks-missed=loop-vectorize < %s 2>&1 | \
|
|
; RUN: FileCheck -check-prefix=NO_HOTNESS -check-prefix=BOTH %s
|
|
|
|
|
|
; 1 void cold(char *A, char *B, char *C, char *D, char *E, int N) {
|
|
; 2 for(int i = 0; i < N; i++) {
|
|
; 3 A[i + 1] = A[i] + B[i];
|
|
; 4 C[i] = Dptr E[i];
|
|
; 5 }
|
|
; 6 }
|
|
; 7
|
|
; 8 void hot(char *A, char *B, char *C, char *D, char *E, int N) {
|
|
; 9 for(int i = 0; i < N; i++) {
|
|
; 10 A[i + 1] = A[i] + B[i];
|
|
; 11 C[i] = Dptr E[i];
|
|
; 12 }
|
|
; 13 }
|
|
; 14
|
|
; 15 void unknown(char *A, char *B, char *C, char *D, char *E, int N) {
|
|
; 16 for(int i = 0; i < N; i++) {
|
|
; 17 A[i + 1] = A[i] + B[i];
|
|
; 18 C[i] = Dptr E[i];
|
|
; 19 }
|
|
; 20 }
|
|
|
|
; HOTNESS: remark: /tmp/s.c:2:3: loop not vectorized (hotness: 300)
|
|
; NO_HOTNESS: remark: /tmp/s.c:2:3: loop not vectorized{{$}}
|
|
; HOTNESS: remark: /tmp/s.c:9:3: loop not vectorized (hotness: 5000)
|
|
; NO_HOTNESS: remark: /tmp/s.c:9:3: loop not vectorized{{$}}
|
|
; BOTH: remark: /tmp/s.c:16:3: loop not vectorized{{$}}
|
|
|
|
; ModuleID = '/tmp/s.c'
|
|
source_filename = "/tmp/s.c"
|
|
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
define void @cold(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture %C, ptr nocapture readonly %D, ptr nocapture readonly %E, i32 %N) #0 !dbg !7 !prof !56 {
|
|
entry:
|
|
%cmp28 = icmp sgt i32 %N, 0, !dbg !9
|
|
br i1 %cmp28, label %ph, label %for.cond.cleanup, !dbg !10, !prof !58
|
|
|
|
ph:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]
|
|
%arrayidx = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !12
|
|
%0 = load i8, ptr %arrayidx, align 1, !dbg !12
|
|
%arrayidx2 = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !16
|
|
%1 = load i8, ptr %arrayidx2, align 1, !dbg !16
|
|
%add = add i8 %1, %0, !dbg !17
|
|
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !10
|
|
%arrayidx7 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv.next, !dbg !18
|
|
store i8 %add, ptr %arrayidx7, align 1, !dbg !19
|
|
%arrayidx9 = getelementptr inbounds i8, ptr %D, i64 %indvars.iv, !dbg !20
|
|
%2 = load i8, ptr %arrayidx9, align 1, !dbg !20
|
|
%arrayidx12 = getelementptr inbounds i8, ptr %E, i64 %indvars.iv, !dbg !21
|
|
%3 = load i8, ptr %arrayidx12, align 1, !dbg !21
|
|
%mul = mul i8 %3, %2, !dbg !22
|
|
%arrayidx16 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !23
|
|
store i8 %mul, ptr %arrayidx16, align 1, !dbg !24
|
|
%lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !10
|
|
%exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !10
|
|
br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !10, !llvm.loop !25, !prof !59
|
|
|
|
for.cond.cleanup:
|
|
ret void, !dbg !11
|
|
}
|
|
|
|
define void @hot(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture %C, ptr nocapture readonly %D, ptr nocapture readonly %E, i32 %N) #0 !dbg !26 !prof !57 {
|
|
entry:
|
|
%cmp28 = icmp sgt i32 %N, 0, !dbg !27
|
|
br i1 %cmp28, label %ph, label %for.cond.cleanup, !dbg !28, !prof !58
|
|
|
|
ph:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]
|
|
%arrayidx = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !30
|
|
%0 = load i8, ptr %arrayidx, align 1, !dbg !30
|
|
%arrayidx2 = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !31
|
|
%1 = load i8, ptr %arrayidx2, align 1, !dbg !31
|
|
%add = add i8 %1, %0, !dbg !32
|
|
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !28
|
|
%arrayidx7 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv.next, !dbg !33
|
|
store i8 %add, ptr %arrayidx7, align 1, !dbg !34
|
|
%arrayidx9 = getelementptr inbounds i8, ptr %D, i64 %indvars.iv, !dbg !35
|
|
%2 = load i8, ptr %arrayidx9, align 1, !dbg !35
|
|
%arrayidx12 = getelementptr inbounds i8, ptr %E, i64 %indvars.iv, !dbg !36
|
|
%3 = load i8, ptr %arrayidx12, align 1, !dbg !36
|
|
%mul = mul i8 %3, %2, !dbg !37
|
|
%arrayidx16 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !38
|
|
store i8 %mul, ptr %arrayidx16, align 1, !dbg !39
|
|
%lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !28
|
|
%exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !28
|
|
br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !28, !llvm.loop !40, !prof !59
|
|
|
|
for.cond.cleanup:
|
|
ret void, !dbg !29
|
|
}
|
|
|
|
define void @unknown(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture %C, ptr nocapture readonly %D, ptr nocapture readonly %E, i32 %N) #0 !dbg !41 {
|
|
entry:
|
|
%cmp28 = icmp sgt i32 %N, 0, !dbg !42
|
|
br i1 %cmp28, label %for.body, label %for.cond.cleanup, !dbg !43
|
|
|
|
for.cond.cleanup:
|
|
ret void, !dbg !44
|
|
|
|
for.body:
|
|
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
|
|
%arrayidx = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !45
|
|
%0 = load i8, ptr %arrayidx, align 1, !dbg !45
|
|
%arrayidx2 = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !46
|
|
%1 = load i8, ptr %arrayidx2, align 1, !dbg !46
|
|
%add = add i8 %1, %0, !dbg !47
|
|
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !43
|
|
%arrayidx7 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv.next, !dbg !48
|
|
store i8 %add, ptr %arrayidx7, align 1, !dbg !49
|
|
%arrayidx9 = getelementptr inbounds i8, ptr %D, i64 %indvars.iv, !dbg !50
|
|
%2 = load i8, ptr %arrayidx9, align 1, !dbg !50
|
|
%arrayidx12 = getelementptr inbounds i8, ptr %E, i64 %indvars.iv, !dbg !51
|
|
%3 = load i8, ptr %arrayidx12, align 1, !dbg !51
|
|
%mul = mul i8 %3, %2, !dbg !52
|
|
%arrayidx16 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !53
|
|
store i8 %mul, ptr %arrayidx16, align 1, !dbg !54
|
|
%lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !43
|
|
%exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !43
|
|
br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !43, !llvm.loop !55
|
|
}
|
|
|
|
attributes #0 = { "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "target-features"="+cx16,+fxsr,+mmx,+sse,+sse2,+sse3,+ssse3,+x87" "use-soft-float"="false" }
|
|
|
|
!llvm.dbg.cu = !{!0}
|
|
!llvm.module.flags = !{!3, !4, !5}
|
|
|
|
!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.9.0 (trunk 273572) (llvm/trunk 273585)", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2)
|
|
!1 = !DIFile(filename: "/tmp/s.c", directory: "/tmp")
|
|
!2 = !{}
|
|
!3 = !{i32 2, !"Dwarf Version", i32 2}
|
|
!4 = !{i32 2, !"Debug Info Version", i32 3}
|
|
!5 = !{i32 1, !"PIC Level", i32 2}
|
|
!7 = distinct !DISubprogram(name: "cold", scope: !1, file: !1, line: 1, type: !8, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)
|
|
!8 = !DISubroutineType(types: !2)
|
|
!9 = !DILocation(line: 2, column: 20, scope: !7)
|
|
!10 = !DILocation(line: 2, column: 3, scope: !7)
|
|
!11 = !DILocation(line: 6, column: 1, scope: !7)
|
|
!12 = !DILocation(line: 3, column: 16, scope: !7)
|
|
!16 = !DILocation(line: 3, column: 23, scope: !7)
|
|
!17 = !DILocation(line: 3, column: 21, scope: !7)
|
|
!18 = !DILocation(line: 3, column: 5, scope: !7)
|
|
!19 = !DILocation(line: 3, column: 14, scope: !7)
|
|
!20 = !DILocation(line: 4, column: 12, scope: !7)
|
|
!21 = !DILocation(line: 4, column: 19, scope: !7)
|
|
!22 = !DILocation(line: 4, column: 17, scope: !7)
|
|
!23 = !DILocation(line: 4, column: 5, scope: !7)
|
|
!24 = !DILocation(line: 4, column: 10, scope: !7)
|
|
!25 = distinct !{!25, !10}
|
|
!26 = distinct !DISubprogram(name: "hot", scope: !1, file: !1, line: 8, type: !8, isLocal: false, isDefinition: true, scopeLine: 8, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)
|
|
!27 = !DILocation(line: 9, column: 20, scope: !26)
|
|
!28 = !DILocation(line: 9, column: 3, scope: !26)
|
|
!29 = !DILocation(line: 13, column: 1, scope: !26)
|
|
!30 = !DILocation(line: 10, column: 16, scope: !26)
|
|
!31 = !DILocation(line: 10, column: 23, scope: !26)
|
|
!32 = !DILocation(line: 10, column: 21, scope: !26)
|
|
!33 = !DILocation(line: 10, column: 5, scope: !26)
|
|
!34 = !DILocation(line: 10, column: 14, scope: !26)
|
|
!35 = !DILocation(line: 11, column: 12, scope: !26)
|
|
!36 = !DILocation(line: 11, column: 19, scope: !26)
|
|
!37 = !DILocation(line: 11, column: 17, scope: !26)
|
|
!38 = !DILocation(line: 11, column: 5, scope: !26)
|
|
!39 = !DILocation(line: 11, column: 10, scope: !26)
|
|
!40 = distinct !{!40, !28}
|
|
!41 = distinct !DISubprogram(name: "unknown", scope: !1, file: !1, line: 15, type: !8, isLocal: false, isDefinition: true, scopeLine: 15, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)
|
|
!42 = !DILocation(line: 16, column: 20, scope: !41)
|
|
!43 = !DILocation(line: 16, column: 3, scope: !41)
|
|
!44 = !DILocation(line: 20, column: 1, scope: !41)
|
|
!45 = !DILocation(line: 17, column: 16, scope: !41)
|
|
!46 = !DILocation(line: 17, column: 23, scope: !41)
|
|
!47 = !DILocation(line: 17, column: 21, scope: !41)
|
|
!48 = !DILocation(line: 17, column: 5, scope: !41)
|
|
!49 = !DILocation(line: 17, column: 14, scope: !41)
|
|
!50 = !DILocation(line: 18, column: 12, scope: !41)
|
|
!51 = !DILocation(line: 18, column: 19, scope: !41)
|
|
!52 = !DILocation(line: 18, column: 17, scope: !41)
|
|
!53 = !DILocation(line: 18, column: 5, scope: !41)
|
|
!54 = !DILocation(line: 18, column: 10, scope: !41)
|
|
!55 = distinct !{!55, !43}
|
|
!56 = !{!"function_entry_count", i64 3}
|
|
!57 = !{!"function_entry_count", i64 50}
|
|
!58 = !{!"branch_weights", i32 10000, i32 1}
|
|
!59 = !{!"branch_weights", i32 1, i32 99}
|