1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -basic-aa -slp-vectorizer -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 3 4target 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" 5target triple = "x86_64-apple-macosx10.8.0" 6 7; Simple 3-pair chain with loads and stores 8define void @test1(double* %a, double* %b, double* %c) { 9; CHECK-LABEL: @test1( 10; CHECK-NEXT: entry: 11; CHECK-NEXT: [[AGG_TMP_I_I_SROA_0:%.*]] = alloca [3 x double], align 16 12; CHECK-NEXT: [[STORE1:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[AGG_TMP_I_I_SROA_0]], i64 0, i64 1 13; CHECK-NEXT: [[STORE2:%.*]] = getelementptr inbounds [3 x double], [3 x double]* [[AGG_TMP_I_I_SROA_0]], i64 0, i64 2 14; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds double, double* [[A:%.*]], i64 1 15; CHECK-NEXT: [[TMP0:%.*]] = bitcast double* [[A]] to <2 x double>* 16; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* [[TMP0]], align 8 17; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds double, double* [[B:%.*]], i64 1 18; CHECK-NEXT: [[TMP2:%.*]] = bitcast double* [[B]] to <2 x double>* 19; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* [[TMP2]], align 8 20; CHECK-NEXT: [[TMP4:%.*]] = fmul <2 x double> [[TMP1]], [[TMP3]] 21; CHECK-NEXT: [[TMP5:%.*]] = bitcast double* [[STORE1]] to <2 x double>* 22; CHECK-NEXT: store <2 x double> [[TMP4]], <2 x double>* [[TMP5]], align 8 23; CHECK-NEXT: ret void 24; 25entry: 26 %agg.tmp.i.i.sroa.0 = alloca [3 x double], align 16 27 %i0 = load double, double* %a 28 %i1 = load double, double* %b 29 %mul = fmul double %i0, %i1 30 %store1 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 1 31 %store2 = getelementptr inbounds [3 x double], [3 x double]* %agg.tmp.i.i.sroa.0, i64 0, i64 2 32 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 33 %i3 = load double, double* %arrayidx3, align 8 34 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 35 %i4 = load double, double* %arrayidx4, align 8 36 %mul5 = fmul double %i3, %i4 37 store double %mul, double* %store1 38 store double %mul5, double* %store2, align 16 39 ret void 40} 41 42; Float has 4 byte abi alignment on x86_64. We must use the alignment of the 43; value being loaded/stored not the alignment of the pointer type. 44 45define void @test2(float * %a, float * %b) { 46; CHECK-LABEL: @test2( 47; CHECK-NEXT: entry: 48; CHECK-NEXT: [[A1:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 1 49; CHECK-NEXT: [[A2:%.*]] = getelementptr inbounds float, float* [[A]], i64 2 50; CHECK-NEXT: [[A3:%.*]] = getelementptr inbounds float, float* [[A]], i64 3 51; CHECK-NEXT: [[TMP0:%.*]] = bitcast float* [[A]] to <4 x float>* 52; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* [[TMP0]], align 4 53; CHECK-NEXT: [[B1:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 1 54; CHECK-NEXT: [[B2:%.*]] = getelementptr inbounds float, float* [[B]], i64 2 55; CHECK-NEXT: [[B3:%.*]] = getelementptr inbounds float, float* [[B]], i64 3 56; CHECK-NEXT: [[TMP2:%.*]] = bitcast float* [[B]] to <4 x float>* 57; CHECK-NEXT: store <4 x float> [[TMP1]], <4 x float>* [[TMP2]], align 4 58; CHECK-NEXT: ret void 59; 60entry: 61 %l0 = load float, float* %a 62 %a1 = getelementptr inbounds float, float* %a, i64 1 63 %l1 = load float, float* %a1 64 %a2 = getelementptr inbounds float, float* %a, i64 2 65 %l2 = load float, float* %a2 66 %a3 = getelementptr inbounds float, float* %a, i64 3 67 %l3 = load float, float* %a3 68 store float %l0, float* %b 69 %b1 = getelementptr inbounds float, float* %b, i64 1 70 store float %l1, float* %b1 71 %b2 = getelementptr inbounds float, float* %b, i64 2 72 store float %l2, float* %b2 73 %b3 = getelementptr inbounds float, float* %b, i64 3 74 store float %l3, float* %b3 75 ret void 76} 77