1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -instcombine -S | FileCheck %s
define i1 @reduce_xor_self(<8 x i1> %x) {
; CHECK-LABEL: @reduce_xor_self(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <8 x i1> [[X:%.*]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = call i8 @llvm.ctpop.i8(i8 [[TMP1]]), !range [[RNG0:![0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = and i8 [[TMP2]], 1
; CHECK-NEXT: [[TMP4:%.*]] = icmp ne i8 [[TMP3]], 0
; CHECK-NEXT: ret i1 [[TMP4]]
;
%res = call i1 @llvm.vector.reduce.xor.v8i32(<8 x i1> %x)
ret i1 %res
}
define i32 @reduce_xor_sext(<4 x i1> %x) {
; CHECK-LABEL: @reduce_xor_sext(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i1> [[X:%.*]] to i4
; CHECK-NEXT: [[TMP2:%.*]] = call i4 @llvm.ctpop.i4(i4 [[TMP1]]), !range [[RNG1:![0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = and i4 [[TMP2]], 1
; CHECK-NEXT: [[SEXT:%.*]] = sub nsw i4 0, [[TMP3]]
; CHECK-NEXT: [[TMP4:%.*]] = sext i4 [[SEXT]] to i32
; CHECK-NEXT: ret i32 [[TMP4]]
;
%sext = sext <4 x i1> %x to <4 x i32>
%res = call i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> %sext)
ret i32 %res
}
define i64 @reduce_xor_zext(<8 x i1> %x) {
; CHECK-LABEL: @reduce_xor_zext(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <8 x i1> [[X:%.*]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = call i8 @llvm.ctpop.i8(i8 [[TMP1]]), !range [[RNG0]]
; CHECK-NEXT: [[TMP3:%.*]] = and i8 [[TMP2]], 1
; CHECK-NEXT: [[TMP4:%.*]] = zext i8 [[TMP3]] to i64
; CHECK-NEXT: ret i64 [[TMP4]]
;
%zext = zext <8 x i1> %x to <8 x i64>
%res = call i64 @llvm.vector.reduce.xor.v8i64(<8 x i64> %zext)
ret i64 %res
}
define i16 @reduce_xor_sext_same(<16 x i1> %x) {
; CHECK-LABEL: @reduce_xor_sext_same(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <16 x i1> [[X:%.*]] to i16
; CHECK-NEXT: [[TMP2:%.*]] = call i16 @llvm.ctpop.i16(i16 [[TMP1]]), !range [[RNG2:![0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = and i16 [[TMP2]], 1
; CHECK-NEXT: [[SEXT:%.*]] = sub nsw i16 0, [[TMP3]]
; CHECK-NEXT: ret i16 [[SEXT]]
;
%sext = sext <16 x i1> %x to <16 x i16>
%res = call i16 @llvm.vector.reduce.xor.v16i16(<16 x i16> %sext)
ret i16 %res
}
define i8 @reduce_xor_zext_long(<128 x i1> %x) {
; CHECK-LABEL: @reduce_xor_zext_long(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <128 x i1> [[X:%.*]] to i128
; CHECK-NEXT: [[TMP2:%.*]] = call i128 @llvm.ctpop.i128(i128 [[TMP1]]), !range [[RNG3:![0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = trunc i128 [[TMP2]] to i8
; CHECK-NEXT: [[TMP4:%.*]] = and i8 [[TMP3]], 1
; CHECK-NEXT: [[TMP5:%.*]] = sub nsw i8 0, [[TMP4]]
; CHECK-NEXT: ret i8 [[TMP5]]
;
%sext = sext <128 x i1> %x to <128 x i8>
%res = call i8 @llvm.vector.reduce.xor.v128i8(<128 x i8> %sext)
ret i8 %res
}
@glob = external global i8, align 1
define i8 @reduce_xor_zext_long_external_use(<128 x i1> %x) {
; CHECK-LABEL: @reduce_xor_zext_long_external_use(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <128 x i1> [[X:%.*]] to i128
; CHECK-NEXT: [[TMP2:%.*]] = call i128 @llvm.ctpop.i128(i128 [[TMP1]]), !range [[RNG3]]
; CHECK-NEXT: [[TMP3:%.*]] = trunc i128 [[TMP2]] to i8
; CHECK-NEXT: [[TMP4:%.*]] = and i8 [[TMP3]], 1
; CHECK-NEXT: [[TMP5:%.*]] = sub nsw i8 0, [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = extractelement <128 x i1> [[X]], i64 0
; CHECK-NEXT: [[EXT:%.*]] = sext i1 [[TMP6]] to i8
; CHECK-NEXT: store i8 [[EXT]], i8* @glob, align 1
; CHECK-NEXT: ret i8 [[TMP5]]
;
%sext = sext <128 x i1> %x to <128 x i8>
%res = call i8 @llvm.vector.reduce.xor.v128i8(<128 x i8> %sext)
%ext = extractelement <128 x i8> %sext, i32 0
store i8 %ext, i8* @glob, align 1
ret i8 %res
}
@glob1 = external global i64, align 8
define i64 @reduce_xor_zext_external_use(<8 x i1> %x) {
; CHECK-LABEL: @reduce_xor_zext_external_use(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <8 x i1> [[X:%.*]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = call i8 @llvm.ctpop.i8(i8 [[TMP1]]), !range [[RNG0]]
; CHECK-NEXT: [[TMP3:%.*]] = and i8 [[TMP2]], 1
; CHECK-NEXT: [[TMP4:%.*]] = zext i8 [[TMP3]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <8 x i1> [[X]], i64 0
; CHECK-NEXT: [[EXT:%.*]] = zext i1 [[TMP5]] to i64
; CHECK-NEXT: store i64 [[EXT]], i64* @glob1, align 8
; CHECK-NEXT: ret i64 [[TMP4]]
;
%zext = zext <8 x i1> %x to <8 x i64>
%res = call i64 @llvm.vector.reduce.xor.v8i64(<8 x i64> %zext)
%ext = extractelement <8 x i64> %zext, i32 0
store i64 %ext, i64* @glob1, align 8
ret i64 %res
}
declare i1 @llvm.vector.reduce.xor.v8i32(<8 x i1> %a)
declare i32 @llvm.vector.reduce.xor.v4i32(<4 x i32> %a)
declare i64 @llvm.vector.reduce.xor.v8i64(<8 x i64> %a)
declare i16 @llvm.vector.reduce.xor.v16i16(<16 x i16> %a)
declare i8 @llvm.vector.reduce.xor.v128i8(<128 x i8> %a)
|
