1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=slp-vectorizer,dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 3 4target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128" 5target triple = "i386-apple-macosx10.9.0" 6 7; We disable the vectorization of <3 x float> for now 8 9; float foo(ptr A) { 10; 11; float R = A[0]; 12; float G = A[1]; 13; float B = A[2]; 14; for (int i=0; i < 121; i+=3) { 15; R+=Aptr7; 16; G+=Aptr8; 17; B+=Aptr9; 18; } 19; 20; return R+G+B; 21; } 22 23define float @foo(ptr nocapture readonly %A) { 24; CHECK-LABEL: @foo( 25; CHECK-NEXT: entry: 26; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr [[A:%.*]], align 4 27; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 2 28; CHECK-NEXT: [[TMP2:%.*]] = load float, ptr [[ARRAYIDX2]], align 4 29; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x float> [[TMP1]], i32 0 30; CHECK-NEXT: br label [[FOR_BODY:%.*]] 31; CHECK: for.body: 32; CHECK-NEXT: [[TMP4:%.*]] = phi float [ [[TMP3]], [[ENTRY:%.*]] ], [ [[DOTPRE:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE:%.*]] ] 33; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ] 34; CHECK-NEXT: [[B_032:%.*]] = phi float [ [[TMP2]], [[ENTRY]] ], [ [[ADD14:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ] 35; CHECK-NEXT: [[TMP5:%.*]] = phi <2 x float> [ [[TMP1]], [[ENTRY]] ], [ [[TMP11:%.*]], [[FOR_BODY_FOR_BODY_CRIT_EDGE]] ] 36; CHECK-NEXT: [[TMP6:%.*]] = add nsw i64 [[INDVARS_IV]], 1 37; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]] 38; CHECK-NEXT: [[TMP7:%.*]] = load float, ptr [[ARRAYIDX7]], align 4 39; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP4]], i32 0 40; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i32 1 41; CHECK-NEXT: [[TMP10:%.*]] = fmul <2 x float> [[TMP9]], <float 7.000000e+00, float 8.000000e+00> 42; CHECK-NEXT: [[TMP11]] = fadd <2 x float> [[TMP5]], [[TMP10]] 43; CHECK-NEXT: [[TMP12:%.*]] = add nsw i64 [[INDVARS_IV]], 2 44; CHECK-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP12]] 45; CHECK-NEXT: [[TMP13:%.*]] = load float, ptr [[ARRAYIDX12]], align 4 46; CHECK-NEXT: [[MUL13:%.*]] = fmul float [[TMP13]], 9.000000e+00 47; CHECK-NEXT: [[ADD14]] = fadd float [[B_032]], [[MUL13]] 48; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 3 49; CHECK-NEXT: [[TMP14:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32 50; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[TMP14]], 121 51; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY_FOR_BODY_CRIT_EDGE]], label [[FOR_END:%.*]] 52; CHECK: for.body.for.body_crit_edge: 53; CHECK-NEXT: [[ARRAYIDX3_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDVARS_IV_NEXT]] 54; CHECK-NEXT: [[DOTPRE]] = load float, ptr [[ARRAYIDX3_PHI_TRANS_INSERT]], align 4 55; CHECK-NEXT: br label [[FOR_BODY]] 56; CHECK: for.end: 57; CHECK-NEXT: [[TMP15:%.*]] = extractelement <2 x float> [[TMP11]], i32 0 58; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x float> [[TMP11]], i32 1 59; CHECK-NEXT: [[ADD16:%.*]] = fadd float [[TMP15]], [[TMP16]] 60; CHECK-NEXT: [[ADD17:%.*]] = fadd float [[ADD16]], [[ADD14]] 61; CHECK-NEXT: ret float [[ADD17]] 62; 63entry: 64 %0 = load float, ptr %A, align 4 65 %arrayidx1 = getelementptr inbounds float, ptr %A, i64 1 66 %1 = load float, ptr %arrayidx1, align 4 67 %arrayidx2 = getelementptr inbounds float, ptr %A, i64 2 68 %2 = load float, ptr %arrayidx2, align 4 69 br label %for.body 70 71for.body: ; preds = %for.body.for.body_crit_edge, %entry 72 %3 = phi float [ %0, %entry ], [ %.pre, %for.body.for.body_crit_edge ] 73 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body.for.body_crit_edge ] 74 %B.032 = phi float [ %2, %entry ], [ %add14, %for.body.for.body_crit_edge ] 75 %G.031 = phi float [ %1, %entry ], [ %add9, %for.body.for.body_crit_edge ] 76 %R.030 = phi float [ %0, %entry ], [ %add4, %for.body.for.body_crit_edge ] 77 %mul = fmul float %3, 7.000000e+00 78 %add4 = fadd float %R.030, %mul 79 %4 = add nsw i64 %indvars.iv, 1 80 %arrayidx7 = getelementptr inbounds float, ptr %A, i64 %4 81 %5 = load float, ptr %arrayidx7, align 4 82 %mul8 = fmul float %5, 8.000000e+00 83 %add9 = fadd float %G.031, %mul8 84 %6 = add nsw i64 %indvars.iv, 2 85 %arrayidx12 = getelementptr inbounds float, ptr %A, i64 %6 86 %7 = load float, ptr %arrayidx12, align 4 87 %mul13 = fmul float %7, 9.000000e+00 88 %add14 = fadd float %B.032, %mul13 89 %indvars.iv.next = add i64 %indvars.iv, 3 90 %8 = trunc i64 %indvars.iv.next to i32 91 %cmp = icmp slt i32 %8, 121 92 br i1 %cmp, label %for.body.for.body_crit_edge, label %for.end 93 94for.body.for.body_crit_edge: ; preds = %for.body 95 %arrayidx3.phi.trans.insert = getelementptr inbounds float, ptr %A, i64 %indvars.iv.next 96 %.pre = load float, ptr %arrayidx3.phi.trans.insert, align 4 97 br label %for.body 98 99for.end: ; preds = %for.body 100 %add16 = fadd float %add4, %add9 101 %add17 = fadd float %add16, %add14 102 ret float %add17 103} 104 105