1; RUN: opt -mtriple=thumbv8.1m.main-arm-eabihf < %s -loop-vectorize -S | \
2; RUN:  FileCheck %s -check-prefixes=CHECK,NO-FOLDING
3
4; RUN: opt -mtriple=thumbv8.1m.main-arm-eabihf -mattr=-mve < %s -loop-vectorize -enable-arm-maskedldst=true -S | \
5; RUN:  FileCheck %s -check-prefixes=CHECK,NO-FOLDING
6
7; RUN: opt -mtriple=thumbv8.1m.main-arm-eabihf -mattr=+mve < %s -loop-vectorize -enable-arm-maskedldst=false -S | \
8; RUN:  FileCheck %s -check-prefixes=CHECK,NO-FOLDING
9
10; Disabling the low-overhead branch extension will make
11; 'isHardwareLoopProfitable' return false, so that we test avoiding folding for
12; these cases.
13; RUN: opt -mtriple=thumbv8.1m.main-arm-eabihf -mattr=+mve,-lob < %s -loop-vectorize -enable-arm-maskedldst=true -S | \
14; RUN:  FileCheck %s -check-prefixes=CHECK,NO-FOLDING
15
16; RUN: opt -mtriple=thumbv8.1m.main-arm-eabihf -mattr=+mve < %s -loop-vectorize -enable-arm-maskedldst=true -S | \
17; RUN:  FileCheck %s -check-prefixes=CHECK,PREFER-FOLDING
18
19define dso_local void @tail_folding(i32* noalias nocapture %A, i32* noalias nocapture readonly %B, i32* noalias nocapture readonly %C) {
20; CHECK-LABEL: tail_folding(
21;
22; NO-FOLDING-NOT:  call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(
23; NO-FOLDING-NOT:  call void @llvm.masked.store.v4i32.p0v4i32(
24;
25; TODO: this needs implementation of TTI::preferPredicateOverEpilogue,
26; then this will be tail-folded too:
27;
28; PREFER-FOLDING-NOT:  call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(
29; PREFER-FOLDING-NOT:  call void @llvm.masked.store.v4i32.p0v4i32(
30;
31entry:
32  br label %for.body
33
34for.cond.cleanup:
35  ret void
36
37for.body:
38  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
39  %arrayidx = getelementptr inbounds i32, i32* %B, i64 %indvars.iv
40  %0 = load i32, i32* %arrayidx, align 4
41  %arrayidx2 = getelementptr inbounds i32, i32* %C, i64 %indvars.iv
42  %1 = load i32, i32* %arrayidx2, align 4
43  %add = add nsw i32 %1, %0
44  %arrayidx4 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
45  store i32 %add, i32* %arrayidx4, align 4
46  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
47  %exitcond = icmp eq i64 %indvars.iv.next, 430
48  br i1 %exitcond, label %for.cond.cleanup, label %for.body
49}
50