40 lines · plain
1; RUN: opt -S -passes=loop-vectorize,instcombine -force-vector-width=4 < %s | FileCheck %s2 3target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"4 5; Vectorization of loop with bitcast between GEP and load6; Simplified source code:7;void foo (ptr __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: @foo14; CHECK: vector.body15; CHECK: %[[IV:.+]] = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]16; CHECK: %[[v0:.+]] = getelementptr inbounds ptr, ptr %in, i64 %[[IV]]17; CHECK: %wide.load = load <4 x i64>, ptr %[[v0]], align 818; CHECK: icmp eq <4 x i64> %wide.load, zeroinitializer19; CHECK: br i120 21define void @foo(ptr noalias nocapture readonly %in, ptr noalias nocapture readnone %out, ptr noalias nocapture %res) #0 {22entry:23 br label %for.body24 25for.body:26 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]27 %arrayidx = getelementptr inbounds ptr, ptr %in, i64 %indvars.iv28 %tmp54 = load i64, ptr %arrayidx, align 829 %cmp1 = icmp eq i64 %tmp54, 030 %arrayidx3 = getelementptr inbounds i8, ptr %res, i64 %indvars.iv31 %frombool = zext i1 %cmp1 to i832 store i8 %frombool, ptr %arrayidx3, align 133 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 134 %exitcond = icmp eq i64 %indvars.iv.next, 409635 br i1 %exitcond, label %for.end, label %for.body36 37for.end:38 ret void39}40