1 ; RUN: opt < %s -loop-accesses -analyze | FileCheck -check-prefix=OLDPM %s
2 ; RUN: opt -passes='require<scalar-evolution>,require<aa>,loop(print-access-info)' -disable-output < %s 2>&1 | FileCheck -check-prefix=NEWPM %s
4 ; Test to confirm LAA will find multiple stores to an invariant address in the
7 ; for(; i < itr; i++) {
8 ; for(; j < itr; j++) {
9 ; var1[i] = var2[j] + var1[i];
14 ; The LAA with the new PM is a loop pass so we go from inner to outer loops.
16 ; OLDPM: for.cond1.preheader:
17 ; OLDPM: Non vectorizable stores to invariant address were not found in loop.
19 ; OLDPM: Non vectorizable stores to invariant address were found in loop.
22 ; NEWPM: Non vectorizable stores to invariant address were found in loop.
23 ; NEWPM: for.cond1.preheader:
24 ; NEWPM: Non vectorizable stores to invariant address were not found in loop.
26 define i32 @foo(i32* nocapture %var1, i32* nocapture readonly %var2, i32 %itr) #0 {
28 %cmp20 = icmp eq i32 %itr, 0
29 br i1 %cmp20, label %for.end10, label %for.cond1.preheader
31 for.cond1.preheader: ; preds = %entry, %for.inc8
32 %indvars.iv23 = phi i64 [ %indvars.iv.next24, %for.inc8 ], [ 0, %entry ]
33 %j.022 = phi i32 [ %j.1.lcssa, %for.inc8 ], [ 0, %entry ]
34 %cmp218 = icmp ult i32 %j.022, %itr
35 br i1 %cmp218, label %for.body3.lr.ph, label %for.inc8
37 for.body3.lr.ph: ; preds = %for.cond1.preheader
38 %arrayidx5 = getelementptr inbounds i32, i32* %var1, i64 %indvars.iv23
39 %0 = zext i32 %j.022 to i64
42 for.body3: ; preds = %for.body3, %for.body3.lr.ph
43 %indvars.iv = phi i64 [ %0, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]
44 %arrayidx = getelementptr inbounds i32, i32* %var2, i64 %indvars.iv
45 %1 = load i32, i32* %arrayidx, align 4
46 %2 = load i32, i32* %arrayidx5, align 4
47 %add = add nsw i32 %2, %1
48 store i32 %add, i32* %arrayidx5, align 4
49 %3 = load i32, i32* %arrayidx5, align 4
50 %4 = add nsw i32 %3, 1
51 store i32 %4, i32* %arrayidx5, align 4
52 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
53 %lftr.wideiv = trunc i64 %indvars.iv.next to i32
54 %exitcond = icmp eq i32 %lftr.wideiv, %itr
55 br i1 %exitcond, label %for.inc8, label %for.body3
57 for.inc8: ; preds = %for.body3, %for.cond1.preheader
58 %j.1.lcssa = phi i32 [ %j.022, %for.cond1.preheader ], [ %itr, %for.body3 ]
59 %indvars.iv.next24 = add nuw nsw i64 %indvars.iv23, 1
60 %lftr.wideiv25 = trunc i64 %indvars.iv.next24 to i32
61 %exitcond26 = icmp eq i32 %lftr.wideiv25, %itr
62 br i1 %exitcond26, label %for.end10, label %for.cond1.preheader
64 for.end10: ; preds = %for.inc8, %entry