1; RUN: opt < %s -sroa -S | FileCheck %s 2; 3; Make sure the llvm.access.group meta-data is preserved 4; when a load/store is replaced with another load/store by sroa 5; 6; class Complex { 7; private: 8; float real_; 9; float imaginary_; 10; 11; public: 12; Complex() : real_(0), imaginary_(0) { } 13; Complex(float real, float imaginary) : real_(real), imaginary_(imaginary) { } 14; Complex(const Complex &rhs) : real_(rhs.real()), imaginary_(rhs.imaginary()) { } 15; 16; inline float real() const { return real_; } 17; inline float imaginary() const { return imaginary_; } 18; 19; Complex operator+(const Complex& rhs) const 20; { 21; return Complex(real_ + rhs.real_, imaginary_ + rhs.imaginary_); 22; } 23; }; 24; 25; void test(Complex *out, long size) 26; { 27; #pragma clang loop vectorize(assume_safety) 28; for (long offset = 0; offset < size; ++offset) { 29; Complex t0 = out[offset]; 30; out[offset] = t0 + t0; 31; } 32; } 33 34; CHECK: for.body: 35; CHECK-NOT: store i32 %{{.*}}, i32* %{{.*}}, align 4 36; CHECK: store i32 %{{.*}}, i32* %{{.*}}, align 4, !llvm.access.group !1 37; CHECK-NOT: store i32 %{{.*}}, i32* %{{.*}}, align 4 38; CHECK: store i32 %{{.*}}, i32* %{{.*}}, align 4, !llvm.access.group !1 39; CHECK-NOT: store i32 %{{.*}}, i32* %{{.*}}, align 4 40; CHECK: br label 41 42; ModuleID = '<stdin>' 43source_filename = "mem-par-metadata-sroa1.cpp" 44target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 45target triple = "x86_64-unknown-linux-gnu" 46 47%class.Complex = type { float, float } 48 49; Function Attrs: norecurse nounwind uwtable 50define void @_Z4testP7Complexl(%class.Complex* nocapture %out, i64 %size) local_unnamed_addr #0 { 51entry: 52 %t0 = alloca %class.Complex, align 4 53 %ref.tmp = alloca i64, align 8 54 %tmpcast = bitcast i64* %ref.tmp to %class.Complex* 55 br label %for.cond 56 57for.cond: ; preds = %for.body, %entry 58 %offset.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ] 59 %cmp = icmp slt i64 %offset.0, %size 60 br i1 %cmp, label %for.body, label %for.end 61 62for.body: ; preds = %for.cond 63 %arrayidx = getelementptr inbounds %class.Complex, %class.Complex* %out, i64 %offset.0 64 %real_.i = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 0 65 %real_.i.i = getelementptr inbounds %class.Complex, %class.Complex* %arrayidx, i64 0, i32 0 66 %0 = load float, float* %real_.i.i, align 4, !llvm.access.group !11 67 store float %0, float* %real_.i, align 4, !llvm.access.group !11 68 %imaginary_.i = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 1 69 %imaginary_.i.i = getelementptr inbounds %class.Complex, %class.Complex* %arrayidx, i64 0, i32 1 70 %1 = load float, float* %imaginary_.i.i, align 4, !llvm.access.group !11 71 store float %1, float* %imaginary_.i, align 4, !llvm.access.group !11 72 %arrayidx1 = getelementptr inbounds %class.Complex, %class.Complex* %out, i64 %offset.0 73 %real_.i1 = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 0 74 %2 = load float, float* %real_.i1, align 4, !noalias !3, !llvm.access.group !11 75 %real_2.i = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 0 76 %3 = load float, float* %real_2.i, align 4, !noalias !3, !llvm.access.group !11 77 %add.i = fadd float %2, %3 78 %imaginary_.i2 = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 1 79 %4 = load float, float* %imaginary_.i2, align 4, !noalias !3, !llvm.access.group !11 80 %imaginary_3.i = getelementptr inbounds %class.Complex, %class.Complex* %t0, i64 0, i32 1 81 %5 = load float, float* %imaginary_3.i, align 4, !noalias !3, !llvm.access.group !11 82 %add4.i = fadd float %4, %5 83 %real_.i.i3 = getelementptr inbounds %class.Complex, %class.Complex* %tmpcast, i64 0, i32 0 84 store float %add.i, float* %real_.i.i3, align 4, !alias.scope !3, !llvm.access.group !11 85 %imaginary_.i.i4 = getelementptr inbounds %class.Complex, %class.Complex* %tmpcast, i64 0, i32 1 86 store float %add4.i, float* %imaginary_.i.i4, align 4, !alias.scope !3, !llvm.access.group !11 87 %6 = bitcast %class.Complex* %arrayidx1 to i64* 88 %7 = load i64, i64* %ref.tmp, align 8, !llvm.access.group !11 89 store i64 %7, i64* %6, align 4, !llvm.access.group !11 90 %inc = add nsw i64 %offset.0, 1 91 br label %for.cond, !llvm.loop !1 92 93for.end: ; preds = %for.cond 94 ret void 95} 96 97; Function Attrs: argmemonly nounwind 98declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i1) #1 99 100attributes #0 = { norecurse nounwind uwtable "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" "no-signed-zeros-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" } 101attributes #1 = { argmemonly nounwind } 102 103!llvm.ident = !{!0} 104 105!0 = !{!"clang version 4.0.0 (cfe/trunk 277751)"} 106!1 = distinct !{!1, !2, !{!"llvm.loop.parallel_accesses", !11}} 107!2 = !{!"llvm.loop.vectorize.enable", i1 true} 108!3 = !{!4} 109!4 = distinct !{!4, !5, !"_ZNK7ComplexplERKS_: %agg.result"} 110!5 = distinct !{!5, !"_ZNK7ComplexplERKS_"} 111!11 = distinct !{} 112