1 ; RUN: opt -basicaa -loop-vectorize -force-vector-interleave=1 -S -mcpu=core-avx2 2 3 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 4 target triple = "x86_64-unknown-linux-gnu" 5 6 @b = common global [256 x i32] zeroinitializer, align 16 7 @a = common global [256 x i32] zeroinitializer, align 16 8 9 ; unsigned int a[256], b[256]; 10 ; void foo() { 11 ; for (i = 0; i < 256; i++) { 12 ; if (b[i] > a[i]) 13 ; a[i] = b[i]; 14 ; } 15 ; } 16 17 ; CHECK-LABEL: foo 18 ; CHECK: load <8 x i32> 19 ; CHECK: icmp ugt <8 x i32> 20 ; CHECK: masked.store 21 22 define void @foo() { 23 entry: 24 br label %for.body 25 26 for.body: ; preds = %for.inc, %entry 27 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ] 28 %arrayidx = getelementptr inbounds [256 x i32], [256 x i32]* @b, i64 0, i64 %indvars.iv 29 %0 = load i32, i32* %arrayidx, align 4 30 %arrayidx2 = getelementptr inbounds [256 x i32], [256 x i32]* @a, i64 0, i64 %indvars.iv 31 %1 = load i32, i32* %arrayidx2, align 4 32 %cmp3 = icmp ugt i32 %0, %1 33 br i1 %cmp3, label %if.then, label %for.inc 34 35 if.then: ; preds = %for.body 36 store i32 %0, i32* %arrayidx2, align 4 37 br label %for.inc 38 39 for.inc: ; preds = %for.body, %if.then 40 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 41 %exitcond = icmp eq i64 %indvars.iv.next, 256 42 br i1 %exitcond, label %for.end, label %for.body 43 44 for.end: ; preds = %for.inc 45 ret void 46 } 47
