1; RUN: opt < %s -loop-vectorize -pass-remarks-analysis='loop-vectorize' -mtriple=x86_64-unknown-linux -S 2>&1 | FileCheck %s 2 3; Verify analysis remarks are generated when interleaving is not beneficial. 4; CHECK: remark: vectorization-remarks-profitable.c:5:17: the cost-model indicates that vectorization is not beneficial 5; CHECK: remark: vectorization-remarks-profitable.c:5:17: the cost-model indicates that interleaving is not beneficial and is explicitly disabled or interleave count is set to 1 6; CHECK: remark: vectorization-remarks-profitable.c:12:17: the cost-model indicates that vectorization is not beneficial 7; CHECK: remark: vectorization-remarks-profitable.c:12:17: the cost-model indicates that interleaving is not beneficial 8 9; First loop. 10; #pragma clang loop interleave(disable) unroll(disable) 11; for(int i = 0; i < n; i++) { 12; out[i] = *in[i]; 13; } 14 15; Second loop. 16; #pragma clang loop unroll(disable) 17; for(int i = 0; i < n; i++) { 18; out[i] = *in[i]; 19; } 20 21; ModuleID = 'vectorization-remarks-profitable.ll' 22target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128" 23target triple = "x86_64-apple-macosx10.10.0" 24 25; Function Attrs: nounwind uwtable 26define void @do_not_interleave(float** noalias nocapture readonly %in, float* noalias nocapture %out, i32 %size) #0 !dbg !4 { 27entry: 28 %cmp.4 = icmp eq i32 %size, 0, !dbg !10 29 br i1 %cmp.4, label %for.end, label %for.body.preheader, !dbg !11 30 31for.body.preheader: ; preds = %entry 32 br label %for.body, !dbg !12 33 34for.body: ; preds = %for.body.preheader, %for.body 35 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ] 36 %arrayidx = getelementptr inbounds float*, float** %in, i64 %indvars.iv, !dbg !12 37 %0 = bitcast float** %arrayidx to i32**, !dbg !12 38 %1 = load i32*, i32** %0, align 8, !dbg !12 39 %2 = load i32, i32* %1, align 4, !dbg !13 40 %arrayidx2 = getelementptr inbounds float, float* %out, i64 %indvars.iv, !dbg !14 41 %3 = bitcast float* %arrayidx2 to i32*, !dbg !15 42 store i32 %2, i32* %3, align 4, !dbg !15 43 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !11 44 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !11 45 %exitcond = icmp eq i32 %lftr.wideiv, %size, !dbg !11 46 br i1 %exitcond, label %for.end.loopexit, label %for.body, !dbg !11, !llvm.loop !16 47 48for.end.loopexit: ; preds = %for.body 49 br label %for.end, !dbg !19 50 51for.end: ; preds = %for.end.loopexit, %entry 52 ret void, !dbg !19 53} 54 55; Function Attrs: nounwind uwtable 56define void @interleave_not_profitable(float** noalias nocapture readonly %in, float* noalias nocapture %out, i32 %size) #0 !dbg !6 { 57entry: 58 %cmp.4 = icmp eq i32 %size, 0, !dbg !20 59 br i1 %cmp.4, label %for.end, label %for.body, !dbg !21 60 61for.body: ; preds = %entry, %for.body 62 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] 63 %arrayidx = getelementptr inbounds float*, float** %in, i64 %indvars.iv, !dbg !22 64 %0 = bitcast float** %arrayidx to i32**, !dbg !22 65 %1 = load i32*, i32** %0, align 8, !dbg !22 66 %2 = load i32, i32* %1, align 4, !dbg !23 67 %arrayidx2 = getelementptr inbounds float, float* %out, i64 %indvars.iv, !dbg !24 68 %3 = bitcast float* %arrayidx2 to i32*, !dbg !25 69 store i32 %2, i32* %3, align 4, !dbg !25 70 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !21 71 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !21 72 %exitcond = icmp eq i32 %lftr.wideiv, %size, !dbg !21 73 br i1 %exitcond, label %for.end, label %for.body, !dbg !21, !llvm.loop !26 74 75for.end: ; preds = %for.body, %entry 76 ret void, !dbg !27 77} 78 79attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" } 80 81!llvm.dbg.cu = !{!0} 82!llvm.module.flags = !{!7, !8} 83!llvm.ident = !{!9} 84 85!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.8.0 (trunk 250016)", isOptimized: false, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2) 86!1 = !DIFile(filename: "vectorization-remarks-profitable.c", directory: "") 87!2 = !{} 88!4 = distinct !DISubprogram(name: "do_not_interleave", scope: !1, file: !1, line: 1, type: !5, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: false, unit: !0, variables: !2) 89!5 = !DISubroutineType(types: !2) 90!6 = distinct !DISubprogram(name: "interleave_not_profitable", scope: !1, file: !1, line: 8, type: !5, isLocal: false, isDefinition: true, scopeLine: 8, flags: DIFlagPrototyped, isOptimized: false, unit: !0, variables: !2) 91!7 = !{i32 2, !"Dwarf Version", i32 4} 92!8 = !{i32 2, !"Debug Info Version", i32 3} 93!9 = !{!"clang version 3.8.0 (trunk 250016)"} 94!10 = !DILocation(line: 4, column: 23, scope: !4) 95!11 = !DILocation(line: 4, column: 3, scope: !4) 96!12 = !DILocation(line: 5, column: 17, scope: !4) 97!13 = !DILocation(line: 5, column: 16, scope: !4) 98!14 = !DILocation(line: 5, column: 7, scope: !4) 99!15 = !DILocation(line: 5, column: 14, scope: !4) 100!16 = distinct !{!16, !17, !18} 101!17 = !{!"llvm.loop.interleave.count", i32 1} 102!18 = !{!"llvm.loop.unroll.disable"} 103!19 = !DILocation(line: 6, column: 1, scope: !4) 104!20 = !DILocation(line: 11, column: 23, scope: !6) 105!21 = !DILocation(line: 11, column: 3, scope: !6) 106!22 = !DILocation(line: 12, column: 17, scope: !6) 107!23 = !DILocation(line: 12, column: 16, scope: !6) 108!24 = !DILocation(line: 12, column: 7, scope: !6) 109!25 = !DILocation(line: 12, column: 14, scope: !6) 110!26 = distinct !{!26, !18} 111!27 = !DILocation(line: 13, column: 1, scope: !6) 112 113