1; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 2 3target 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" 4target triple = "i386-apple-macosx10.8.0" 5 6; int foo(double *A, int n, int m) { 7; double sum = 0, v1 = 2, v0 = 3; 8; for (int i=0; i < n; ++i) 9; sum += 7*A[i*2] + 7*A[i*2+1]; 10; return sum; 11; } 12 13;CHECK: reduce 14;CHECK: load <2 x double> 15;CHECK: fmul <2 x double> 16;CHECK: ret 17define i32 @reduce(double* nocapture %A, i32 %n, i32 %m) { 18entry: 19 %cmp13 = icmp sgt i32 %n, 0 20 br i1 %cmp13, label %for.body, label %for.end 21 22for.body: ; preds = %entry, %for.body 23 %i.015 = phi i32 [ %inc, %for.body ], [ 0, %entry ] 24 %sum.014 = phi double [ %add6, %for.body ], [ 0.000000e+00, %entry ] 25 %mul = shl nsw i32 %i.015, 1 26 %arrayidx = getelementptr inbounds double, double* %A, i32 %mul 27 %0 = load double, double* %arrayidx, align 4 28 %mul1 = fmul double %0, 7.000000e+00 29 %add12 = or i32 %mul, 1 30 %arrayidx3 = getelementptr inbounds double, double* %A, i32 %add12 31 %1 = load double, double* %arrayidx3, align 4 32 %mul4 = fmul double %1, 7.000000e+00 33 %add5 = fadd double %mul1, %mul4 34 %add6 = fadd double %sum.014, %add5 35 %inc = add nsw i32 %i.015, 1 36 %exitcond = icmp eq i32 %inc, %n 37 br i1 %exitcond, label %for.cond.for.end_crit_edge, label %for.body 38 39for.cond.for.end_crit_edge: ; preds = %for.body 40 %phitmp = fptosi double %add6 to i32 41 br label %for.end 42 43for.end: ; preds = %for.cond.for.end_crit_edge, %entry 44 %sum.0.lcssa = phi i32 [ %phitmp, %for.cond.for.end_crit_edge ], [ 0, %entry ] 45 ret i32 %sum.0.lcssa 46} 47 48