1 ; RUN: opt -S -loop-vectorize -instcombine -force-vector-width=4 < %s | FileCheck %s 2 3 target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128" 4 5 ; Vectorization of loop with bitcast between GEP and load 6 ; Simplified source code: 7 ;void foo (double** __restrict__ in, bool * __restrict__ res) { 8 ; 9 ; for (int i = 0; i < 4096; ++i) 10 ; res[i] = ((unsigned long long)in[i] == 0); 11 ;} 12 13 ; CHECK-LABEL: @foo 14 ; CHECK: vector.body 15 ; CHECK: %0 = getelementptr inbounds double*, double** %in, i64 %index 16 ; CHECK: %1 = bitcast double** %0 to <4 x i64>* 17 ; CHECK: %wide.load = load <4 x i64>, <4 x i64>* %1, align 8 18 ; CHECK: %2 = icmp eq <4 x i64> %wide.load, zeroinitializer 19 ; CHECK: br i1 20 21 define void @foo(double** noalias nocapture readonly %in, double** noalias nocapture readnone %out, i8* noalias nocapture %res) #0 { 22 entry: 23 br label %for.body 24 25 for.body: 26 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] 27 %arrayidx = getelementptr inbounds double*, double** %in, i64 %indvars.iv 28 %tmp53 = bitcast double** %arrayidx to i64* 29 %tmp54 = load i64, i64* %tmp53, align 8 30 %cmp1 = icmp eq i64 %tmp54, 0 31 %arrayidx3 = getelementptr inbounds i8, i8* %res, i64 %indvars.iv 32 %frombool = zext i1 %cmp1 to i8 33 store i8 %frombool, i8* %arrayidx3, align 1 34 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 35 %exitcond = icmp eq i64 %indvars.iv.next, 4096 36 br i1 %exitcond, label %for.end, label %for.body 37 38 for.end: 39 ret void 40 }
