1 ; RUN: opt < %s -O1 -S -loop-versioning-licm -licm -debug-only=loop-versioning-licm -disable-loop-unrolling -enable-new-pm=0 2>&1 | FileCheck %s
2 ; RUN: opt < %s -S -passes='default<O1>,function(loop-versioning-licm,loop-mssa(licm))' -debug-only=loop-versioning-licm -disable-loop-unrolling 2>&1 | FileCheck %s
5 ; Test to confirm loop is a good candidate for LoopVersioningLICM
6 ; It also confirms invariant moved out of loop.
8 ; CHECK: Loop: Loop at depth 2 containing: %for.body3.us<header><latch><exiting>
9 ; CHECK-NEXT: Loop Versioning found to be beneficial
11 ; CHECK: for.cond1.for.inc17_crit_edge.us.loopexit9: ; preds = %for.body3.us
12 ; CHECK-NEXT: %add14.us.lcssa = phi float [ %add14.us, %for.body3.us ]
13 ; CHECK-NEXT: store float %add14.us.lcssa, float* %arrayidx.us, align 4, !alias.scope !0, !noalias !0
14 ; CHECK-NEXT: br label %for.cond1.for.inc17_crit_edge.us
16 define i32 @foo(float* nocapture %var2, float** nocapture readonly %var3, i32 %itr) #0 {
18 %cmp38 = icmp sgt i32 %itr, 1
19 br i1 %cmp38, label %for.body3.lr.ph.us, label %for.end19
21 for.body3.us: ; preds = %for.body3.us, %for.body3.lr.ph.us
22 %0 = phi float [ %.pre, %for.body3.lr.ph.us ], [ %add14.us, %for.body3.us ]
23 %indvars.iv = phi i64 [ 1, %for.body3.lr.ph.us ], [ %indvars.iv.next, %for.body3.us ]
24 %1 = trunc i64 %indvars.iv to i32
25 %conv.us = sitofp i32 %1 to float
26 %add.us = fadd float %conv.us, %0
27 %arrayidx7.us = getelementptr inbounds float, float* %3, i64 %indvars.iv
28 store float %add.us, float* %arrayidx7.us, align 4
29 %2 = load float, float* %arrayidx.us, align 4
30 %add14.us = fadd float %2, %add.us
31 store float %add14.us, float* %arrayidx.us, align 4
32 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
33 %lftr.wideiv = trunc i64 %indvars.iv.next to i32
34 %exitcond = icmp eq i32 %lftr.wideiv, %itr
35 br i1 %exitcond, label %for.cond1.for.inc17_crit_edge.us, label %for.body3.us
37 for.body3.lr.ph.us: ; preds = %entry, %for.cond1.for.inc17_crit_edge.us
38 %indvars.iv40 = phi i64 [ %indvars.iv.next41, %for.cond1.for.inc17_crit_edge.us ], [ 1, %entry ]
39 %arrayidx.us = getelementptr inbounds float, float* %var2, i64 %indvars.iv40
40 %arrayidx6.us = getelementptr inbounds float*, float** %var3, i64 %indvars.iv40
41 %3 = load float*, float** %arrayidx6.us, align 8
42 %.pre = load float, float* %arrayidx.us, align 4
43 br label %for.body3.us
45 for.cond1.for.inc17_crit_edge.us: ; preds = %for.body3.us
46 %indvars.iv.next41 = add nuw nsw i64 %indvars.iv40, 1
47 %lftr.wideiv42 = trunc i64 %indvars.iv.next41 to i32
48 %exitcond43 = icmp eq i32 %lftr.wideiv42, %itr
49 br i1 %exitcond43, label %for.end19, label %for.body3.lr.ph.us
51 for.end19: ; preds = %for.cond1.for.inc17_crit_edge.us, %entry