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=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL
6
7; RUN: opt -passes=loop-vectorize \
8; RUN: -force-tail-folding-style=none \
9; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
10; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP
11
12define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %index, i64 %n) {
13; IF-EVL-LABEL: @gather_scatter(
14; IF-EVL-NEXT:  entry:
15; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
16; IF-EVL:       for.body:
17; IF-EVL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
18; IF-EVL-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], i64 [[INDVARS_IV]]
19; IF-EVL-NEXT:    [[TMP0:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8
20; IF-EVL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], i64 [[TMP0]]
21; IF-EVL-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX5]], align 4
22; IF-EVL-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], i64 [[TMP0]]
23; IF-EVL-NEXT:    store float [[TMP1]], ptr [[ARRAYIDX7]], align 4
24; IF-EVL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
25; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N:%.*]]
26; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
27; IF-EVL:       for.end:
28; IF-EVL-NEXT:    ret void
29;
30; NO-VP-LABEL: @gather_scatter(
31; NO-VP-NEXT:  entry:
32; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
33; NO-VP:       for.body:
34; NO-VP-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
35; NO-VP-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], i64 [[INDVARS_IV]]
36; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8
37; NO-VP-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], i64 [[TMP0]]
38; NO-VP-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX5]], align 4
39; NO-VP-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], i64 [[TMP0]]
40; NO-VP-NEXT:    store float [[TMP1]], ptr [[ARRAYIDX7]], align 4
41; NO-VP-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
42; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N:%.*]]
43; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
44; NO-VP:       for.end:
45; NO-VP-NEXT:    ret void
46;
47entry:
48  br label %for.body
49
50for.body:
51  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
52  %arrayidx3 = getelementptr inbounds i32, ptr %index, i64 %indvars.iv
53  %0 = load i64, ptr %arrayidx3, align 8
54  %arrayidx5 = getelementptr inbounds float, ptr %in, i64 %0
55  %1 = load float, ptr %arrayidx5, align 4
56  %arrayidx7 = getelementptr inbounds float, ptr %out, i64 %0
57  store float %1, ptr %arrayidx7, align 4
58  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
59  %exitcond.not = icmp eq i64 %indvars.iv.next, %n
60  br i1 %exitcond.not, label %for.end, label %for.body
61
62for.end:
63  ret void
64}
65