1; RUN: opt < %s -basicaa -slp-vectorizer -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-LABEL: @test1 8define void @test1(double* %a, double* %b, double* %c) { 9entry: 10 %agg.tmp.i.i.sroa.0 = alloca [3 x double], align 16 11; CHECK: %[[V0:[0-9]+]] = load <2 x double>, <2 x double>* %[[V2:[0-9]+]], align 8 12 %i0 = load double, double* %a 13 %i1 = load double, double* %b 14 %mul = fmul double %i0, %i1 15 %store1 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 1 16 %store2 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 2 17 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 18 %i3 = load double, double* %arrayidx3, align 8 19 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 20 %i4 = load double, double* %arrayidx4, align 8 21 %mul5 = fmul double %i3, %i4 22; CHECK: store <2 x double> %[[V1:[0-9]+]], <2 x double>* %[[V2:[0-9]+]], align 8 23 store double %mul, double* %store1 24 store double %mul5, double* %store2, align 16 25; CHECK: ret 26 ret void 27} 28 29; Float has 4 byte abi alignment on x86_64. We must use the alignmnet of the 30; value being loaded/stored not the alignment of the pointer type. 31 32; CHECK-LABEL: @test2 33; CHECK-NOT: align 8 34; CHECK: load <4 x float>{{.*}}, align 4 35; CHECK: store <4 x float>{{.*}}, align 4 36; CHECK: ret 37 38define void @test2(float * %a, float * %b) { 39entry: 40 %l0 = load float, float* %a 41 %a1 = getelementptr inbounds float, float* %a, i64 1 42 %l1 = load float, float* %a1 43 %a2 = getelementptr inbounds float, float* %a, i64 2 44 %l2 = load float, float* %a2 45 %a3 = getelementptr inbounds float, float* %a, i64 3 46 %l3 = load float, float* %a3 47 store float %l0, float* %b 48 %b1 = getelementptr inbounds float, float* %b, i64 1 49 store float %l1, float* %b1 50 %b2 = getelementptr inbounds float, float* %b, i64 2 51 store float %l2, float* %b2 52 %b3 = getelementptr inbounds float, float* %b, i64 3 53 store float %l3, float* %b3 54 ret void 55} 56