1 ; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 2 3 target 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" 4 target triple = "i386-apple-macosx10.9.0" 5 6 ; float foo(float *A) { 7 ; 8 ; float R = A[0]; 9 ; float G = A[1]; 10 ; float B = A[2]; 11 ; for (int i=0; i < 121; i+=3) { 12 ; R+=A[i+0]*7; 13 ; G+=A[i+1]*8; 14 ; B+=A[i+2]*9; 15 ; } 16 ; 17 ; return R+G+B; 18 ; } 19 20 ;CHECK-LABEL: @foo( 21 ;CHECK: br 22 ;CHECK: phi <3 x float> 23 ;CHECK: fmul <3 x float> 24 ;CHECK: fadd <3 x float> 25 ; At the moment we don't sink extractelements. 26 ;CHECK: br 27 ;CHECK: extractelement 28 ;CHECK: extractelement 29 ;CHECK: extractelement 30 ;CHECK: ret 31 32 define float @foo(float* nocapture readonly %A) { 33 entry: 34 %0 = load float* %A, align 4 35 %arrayidx1 = getelementptr inbounds float* %A, i64 1 36 %1 = load float* %arrayidx1, align 4 37 %arrayidx2 = getelementptr inbounds float* %A, i64 2 38 %2 = load float* %arrayidx2, align 4 39 br label %for.body 40 41 for.body: ; preds = %for.body.for.body_crit_edge, %entry 42 %3 = phi float [ %0, %entry ], [ %.pre, %for.body.for.body_crit_edge ] 43 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body.for.body_crit_edge ] 44 %B.032 = phi float [ %2, %entry ], [ %add14, %for.body.for.body_crit_edge ] 45 %G.031 = phi float [ %1, %entry ], [ %add9, %for.body.for.body_crit_edge ] 46 %R.030 = phi float [ %0, %entry ], [ %add4, %for.body.for.body_crit_edge ] 47 %mul = fmul float %3, 7.000000e+00 48 %add4 = fadd float %R.030, %mul 49 %4 = add nsw i64 %indvars.iv, 1 50 %arrayidx7 = getelementptr inbounds float* %A, i64 %4 51 %5 = load float* %arrayidx7, align 4 52 %mul8 = fmul float %5, 8.000000e+00 53 %add9 = fadd float %G.031, %mul8 54 %6 = add nsw i64 %indvars.iv, 2 55 %arrayidx12 = getelementptr inbounds float* %A, i64 %6 56 %7 = load float* %arrayidx12, align 4 57 %mul13 = fmul float %7, 9.000000e+00 58 %add14 = fadd float %B.032, %mul13 59 %indvars.iv.next = add i64 %indvars.iv, 3 60 %8 = trunc i64 %indvars.iv.next to i32 61 %cmp = icmp slt i32 %8, 121 62 br i1 %cmp, label %for.body.for.body_crit_edge, label %for.end 63 64 for.body.for.body_crit_edge: ; preds = %for.body 65 %arrayidx3.phi.trans.insert = getelementptr inbounds float* %A, i64 %indvars.iv.next 66 %.pre = load float* %arrayidx3.phi.trans.insert, align 4 67 br label %for.body 68 69 for.end: ; preds = %for.body 70 %add16 = fadd float %add4, %add9 71 %add17 = fadd float %add16, %add14 72 ret float %add17 73 } 74 75
