xref: /llvm-project/llvm/test/Transforms/LoopVectorize/PowerPC/vectorize-force-tail-with-evl.ll (revision 413a66f33984a4d484ac1ff0ba7c0ac39ffa3095)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -passes=loop-vectorize \
3; RUN: -force-tail-folding-style=data-with-evl \
4; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
5; RUN: -mtriple=powerpc64le-unknown-linux-gnu \
6; RUN: -mcpu=pwr10 -S < %s | FileCheck %s
7
8; RUN: opt -passes=loop-vectorize \
9; RUN: -force-tail-folding-style=none \
10; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
11; RUN: -mtriple=powerpc64le-unknown-linux-gnu \
12; RUN: -mcpu=pwr10 -S < %s | FileCheck %s
13
14define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
15; CHECK-LABEL: @foo(
16; CHECK-NEXT:  entry:
17; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
18; CHECK:       for.body:
19; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
20; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[IV]]
21; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
22; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[IV]]
23; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
24; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP0]]
25; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]]
26; CHECK-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX4]], align 4
27; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
28; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]]
29; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]]
30; CHECK:       for.cond.cleanup:
31; CHECK-NEXT:    ret void
32;
33entry:
34  br label %for.body
35
36for.body:
37  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
38  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %iv
39  %0 = load i32, ptr %arrayidx, align 4
40  %arrayidx2 = getelementptr inbounds i32, ptr %c, i64 %iv
41  %1 = load i32, ptr %arrayidx2, align 4
42  %add = add nsw i32 %1, %0
43  %arrayidx4 = getelementptr inbounds i32, ptr %a, i64 %iv
44  store i32 %add, ptr %arrayidx4, align 4
45  %iv.next = add nuw nsw i64 %iv, 1
46  %exitcond.not = icmp eq i64 %iv.next, %N
47  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
48
49for.cond.cleanup:
50  ret void
51}
52