51 lines · plain
1; RUN: opt < %s -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s2 3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"4target triple = "x86_64-unknown-linux-gnu"5 6; The parallel loop has been invalidated by the new memory accesses introduced7; by reg2mem (Loop::isParallel() starts to return false). Ensure the loop is8; now non-vectorizable.9 10;CHECK-NOT: <4 x i32>11define void @parallel_loop(ptr nocapture %a, ptr nocapture %b) nounwind uwtable {12entry:13 %indvars.iv.next.reg2mem = alloca i6414 %indvars.iv.reg2mem = alloca i6415 %"reg2mem alloca point" = bitcast i32 0 to i3216 store i64 0, ptr %indvars.iv.reg2mem17 br label %for.body18 19for.body: ; preds = %for.body.for.body_crit_edge, %entry20 %indvars.iv.reload = load i64, ptr %indvars.iv.reg2mem21 %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv.reload22 %0 = load i32, ptr %arrayidx, align 4, !llvm.access.group !423 %arrayidx2 = getelementptr inbounds i32, ptr %a, i64 %indvars.iv.reload24 %1 = load i32, ptr %arrayidx2, align 4, !llvm.access.group !425 %idxprom3 = sext i32 %1 to i6426 %arrayidx4 = getelementptr inbounds i32, ptr %a, i64 %idxprom327 store i32 %0, ptr %arrayidx4, align 4, !llvm.access.group !428 %indvars.iv.next = add i64 %indvars.iv.reload, 129 ; A new store without the parallel metadata here:30 store i64 %indvars.iv.next, ptr %indvars.iv.next.reg2mem31 %indvars.iv.next.reload1 = load i64, ptr %indvars.iv.next.reg2mem32 %arrayidx6 = getelementptr inbounds i32, ptr %b, i64 %indvars.iv.next.reload133 %2 = load i32, ptr %arrayidx6, align 4, !llvm.access.group !434 store i32 %2, ptr %arrayidx2, align 4, !llvm.access.group !435 %indvars.iv.next.reload = load i64, ptr %indvars.iv.next.reg2mem36 %lftr.wideiv = trunc i64 %indvars.iv.next.reload to i3237 %exitcond = icmp eq i32 %lftr.wideiv, 51238 br i1 %exitcond, label %for.end, label %for.body.for.body_crit_edge, !llvm.loop !339 40for.body.for.body_crit_edge: ; preds = %for.body41 %indvars.iv.next.reload2 = load i64, ptr %indvars.iv.next.reg2mem42 store i64 %indvars.iv.next.reload2, ptr %indvars.iv.reg2mem43 br label %for.body44 45for.end: ; preds = %for.body46 ret void47}48 49!3 = !{!3, !{!"llvm.loop.parallel_accesses", !4}}50!4 = distinct !{}51