1; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 2 3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" 4target triple = "x86_64-apple-macosx10.8.0" 5 6; Simple 3-pair chain with loads and stores 7; CHECK: test1 8; CHECK: store <2 x double> 9; CHECK: ret 10define void @test1(double* %a, double* %b, double* %c) { 11entry: 12 %i0 = load double, double* %a, align 8 13 %i1 = load double, double* %b, align 8 14 %mul = fmul double %i0, %i1 15 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 16 %i3 = load double, double* %arrayidx3, align 8 17 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 18 %i4 = load double, double* %arrayidx4, align 8 19 %mul5 = fmul double %i3, %i4 20 store double %mul, double* %c, align 8 21 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 22 store double %mul5, double* %arrayidx5, align 8 23 ret void 24} 25 26; Simple 3-pair chain with loads and stores, obfuscated with bitcasts 27; CHECK: test2 28; CHECK: store <2 x double> 29; CHECK: ret 30define void @test2(double* %a, double* %b, i8* %e) { 31entry: 32 %i0 = load double, double* %a, align 8 33 %i1 = load double, double* %b, align 8 34 %mul = fmul double %i0, %i1 35 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 36 %i3 = load double, double* %arrayidx3, align 8 37 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 38 %i4 = load double, double* %arrayidx4, align 8 39 %mul5 = fmul double %i3, %i4 40 %c = bitcast i8* %e to double* 41 store double %mul, double* %c, align 8 42 %carrayidx5 = getelementptr inbounds i8, i8* %e, i64 8 43 %arrayidx5 = bitcast i8* %carrayidx5 to double* 44 store double %mul5, double* %arrayidx5, align 8 45 ret void 46} 47 48; Don't vectorize volatile loads. 49; CHECK: test_volatile_load 50; CHECK-NOT: load <2 x double> 51; CHECK: store <2 x double> 52; CHECK: ret 53define void @test_volatile_load(double* %a, double* %b, double* %c) { 54entry: 55 %i0 = load volatile double, double* %a, align 8 56 %i1 = load volatile double, double* %b, align 8 57 %mul = fmul double %i0, %i1 58 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 59 %i3 = load double, double* %arrayidx3, align 8 60 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 61 %i4 = load double, double* %arrayidx4, align 8 62 %mul5 = fmul double %i3, %i4 63 store double %mul, double* %c, align 8 64 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 65 store double %mul5, double* %arrayidx5, align 8 66 ret void 67} 68 69; Don't vectorize volatile stores. 70; CHECK: test_volatile_store 71; CHECK-NOT: store <2 x double> 72; CHECK: ret 73define void @test_volatile_store(double* %a, double* %b, double* %c) { 74entry: 75 %i0 = load double, double* %a, align 8 76 %i1 = load double, double* %b, align 8 77 %mul = fmul double %i0, %i1 78 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 79 %i3 = load double, double* %arrayidx3, align 8 80 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 81 %i4 = load double, double* %arrayidx4, align 8 82 %mul5 = fmul double %i3, %i4 83 store volatile double %mul, double* %c, align 8 84 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 85 store volatile double %mul5, double* %arrayidx5, align 8 86 ret void 87} 88 89 90