xref: /llvm-project/llvm/test/Instrumentation/MemorySanitizer/reduce.ll (revision 38fffa630ee80163dc65e759392ad29798905679)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
2; RUN: opt < %s -msan-check-access-address=0 -msan-track-origins=1 -S -passes='module(msan)' 2>&1 | FileCheck -allow-deprecated-dag-overlap --check-prefix=CHECK %s
3
4target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
5target triple = "x86_64-unknown-linux-gnu"
6
7declare i32 @llvm.vector.reduce.add(<3 x i32>)
8declare i32 @llvm.vector.reduce.and(<3 x i32>)
9declare i32 @llvm.vector.reduce.or(<3 x i32>)
10
11define i32 @reduce_add() sanitize_memory {
12; CHECK-LABEL: define i32 @reduce_add(
13; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
14; CHECK-NEXT:    call void @llvm.donothing()
15; CHECK-NEXT:    [[P:%.*]] = inttoptr i64 0 to ptr
16; CHECK-NEXT:    [[O:%.*]] = load <3 x i32>, ptr [[P]], align 16
17; CHECK-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64
18; CHECK-NEXT:    [[TMP2:%.*]] = xor i64 [[TMP1]], 87960930222080
19; CHECK-NEXT:    [[TMP3:%.*]] = inttoptr i64 [[TMP2]] to ptr
20; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[TMP2]], 17592186044416
21; CHECK-NEXT:    [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
22; CHECK-NEXT:    [[_MSLD:%.*]] = load <3 x i32>, ptr [[TMP3]], align 16
23; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 16
24; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vector.reduce.or.v3i32(<3 x i32> [[_MSLD]])
25; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.vector.reduce.add.v3i32(<3 x i32> [[O]])
26; CHECK-NEXT:    store i32 [[TMP7]], ptr @__msan_retval_tls, align 8
27; CHECK-NEXT:    store i32 [[TMP6]], ptr @__msan_retval_origin_tls, align 4
28; CHECK-NEXT:    ret i32 [[R]]
29;
30  %p = inttoptr i64 0 to ptr
31  %o = load <3 x i32>, ptr %p
32  %r = call i32 @llvm.vector.reduce.add(<3 x i32> %o)
33  ret i32 %r
34}
35
36define i32 @reduce_and() sanitize_memory {
37; CHECK-LABEL: define i32 @reduce_and(
38; CHECK-SAME: ) #[[ATTR0]] {
39; CHECK-NEXT:    call void @llvm.donothing()
40; CHECK-NEXT:    [[P:%.*]] = inttoptr i64 0 to ptr
41; CHECK-NEXT:    [[O:%.*]] = load <3 x i32>, ptr [[P]], align 16
42; CHECK-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64
43; CHECK-NEXT:    [[TMP2:%.*]] = xor i64 [[TMP1]], 87960930222080
44; CHECK-NEXT:    [[TMP3:%.*]] = inttoptr i64 [[TMP2]] to ptr
45; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[TMP2]], 17592186044416
46; CHECK-NEXT:    [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
47; CHECK-NEXT:    [[_MSLD:%.*]] = load <3 x i32>, ptr [[TMP3]], align 16
48; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 16
49; CHECK-NEXT:    [[TMP7:%.*]] = or <3 x i32> [[O]], [[_MSLD]]
50; CHECK-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v3i32(<3 x i32> [[TMP7]])
51; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.or.v3i32(<3 x i32> [[_MSLD]])
52; CHECK-NEXT:    [[TMP10:%.*]] = and i32 [[TMP8]], [[TMP9]]
53; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.vector.reduce.and.v3i32(<3 x i32> [[O]])
54; CHECK-NEXT:    store i32 [[TMP10]], ptr @__msan_retval_tls, align 8
55; CHECK-NEXT:    store i32 [[TMP6]], ptr @__msan_retval_origin_tls, align 4
56; CHECK-NEXT:    ret i32 [[R]]
57;
58  %p = inttoptr i64 0 to ptr
59  %o = load <3 x i32>, ptr %p
60  %r = call i32 @llvm.vector.reduce.and(<3 x i32> %o)
61  ret i32 %r
62}
63
64define i32 @reduce_or() sanitize_memory {
65; CHECK-LABEL: define i32 @reduce_or(
66; CHECK-SAME: ) #[[ATTR0]] {
67; CHECK-NEXT:    call void @llvm.donothing()
68; CHECK-NEXT:    [[P:%.*]] = inttoptr i64 0 to ptr
69; CHECK-NEXT:    [[O:%.*]] = load <3 x i32>, ptr [[P]], align 16
70; CHECK-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[P]] to i64
71; CHECK-NEXT:    [[TMP2:%.*]] = xor i64 [[TMP1]], 87960930222080
72; CHECK-NEXT:    [[TMP3:%.*]] = inttoptr i64 [[TMP2]] to ptr
73; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[TMP2]], 17592186044416
74; CHECK-NEXT:    [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
75; CHECK-NEXT:    [[_MSLD:%.*]] = load <3 x i32>, ptr [[TMP3]], align 16
76; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 16
77; CHECK-NEXT:    [[TMP7:%.*]] = xor <3 x i32> [[O]], splat (i32 -1)
78; CHECK-NEXT:    [[TMP8:%.*]] = or <3 x i32> [[TMP7]], [[_MSLD]]
79; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.and.v3i32(<3 x i32> [[TMP8]])
80; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.or.v3i32(<3 x i32> [[_MSLD]])
81; CHECK-NEXT:    [[TMP11:%.*]] = and i32 [[TMP9]], [[TMP10]]
82; CHECK-NEXT:    [[R:%.*]] = call i32 @llvm.vector.reduce.or.v3i32(<3 x i32> [[O]])
83; CHECK-NEXT:    store i32 [[TMP11]], ptr @__msan_retval_tls, align 8
84; CHECK-NEXT:    store i32 [[TMP6]], ptr @__msan_retval_origin_tls, align 4
85; CHECK-NEXT:    ret i32 [[R]]
86;
87  %p = inttoptr i64 0 to ptr
88  %o = load <3 x i32>, ptr %p
89  %r = call i32 @llvm.vector.reduce.or(<3 x i32> %o)
90  ret i32 %r
91}
92