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
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- -mattr=sse | FileCheck %s --check-prefixes=CHECK,SSE
; RUN: llc < %s -mtriple=x86_64-- -mattr=avx | FileCheck %s --check-prefixes=CHECK,AVX
; With no-wrap:
; (X * Y) == 0 --> (X == 0) || (Y == 0)
; (X * Y) != 0 --> (X != 0) && (Y != 0)
define i1 @mul_nsw_eq0_i8(i8 %x, i8 %y) {
; CHECK-LABEL: mul_nsw_eq0_i8:
; CHECK: # %bb.0:
; CHECK-NEXT: testb %sil, %sil
; CHECK-NEXT: sete %cl
; CHECK-NEXT: testb %dil, %dil
; CHECK-NEXT: sete %al
; CHECK-NEXT: orb %cl, %al
; CHECK-NEXT: retq
%m = mul nsw i8 %x, %y
%r = icmp eq i8 %m, 0
ret i1 %r
}
; negative test - not valid if mul can overflow
define i1 @mul_eq0_i8(i8 %x, i8 %y) {
; CHECK-LABEL: mul_eq0_i8:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: mulb %sil
; CHECK-NEXT: testb %al, %al
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%m = mul i8 %x, %y
%r = icmp eq i8 %m, 0
ret i1 %r
}
; negative test - don't try with minsize
define i1 @mul_nsw_eq0_i8_size(i8 %x, i8 %y) minsize {
; CHECK-LABEL: mul_nsw_eq0_i8_size:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: mulb %sil
; CHECK-NEXT: testb %al, %al
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%m = mul nsw i8 %x, %y
%r = icmp eq i8 %m, 0
ret i1 %r
}
define i1 @mul_nsw_ne0_i16(i16 %x, i16 %y) {
; CHECK-LABEL: mul_nsw_ne0_i16:
; CHECK: # %bb.0:
; CHECK-NEXT: testw %si, %si
; CHECK-NEXT: setne %cl
; CHECK-NEXT: testw %di, %di
; CHECK-NEXT: setne %al
; CHECK-NEXT: andb %cl, %al
; CHECK-NEXT: retq
%m = mul nsw i16 %x, %y
%r = icmp ne i16 %m, 0
ret i1 %r
}
define i1 @mul_nuw_eq0_i32(i32 %x, i32 %y) {
; CHECK-LABEL: mul_nuw_eq0_i32:
; CHECK: # %bb.0:
; CHECK-NEXT: testl %esi, %esi
; CHECK-NEXT: sete %cl
; CHECK-NEXT: testl %edi, %edi
; CHECK-NEXT: sete %al
; CHECK-NEXT: orb %cl, %al
; CHECK-NEXT: retq
%m = mul nuw i32 %x, %y
%r = icmp eq i32 %m, 0
ret i1 %r
}
define i1 @mul_nsw_nuw_ne0_i64(i64 %x, i64 %y) {
; CHECK-LABEL: mul_nsw_nuw_ne0_i64:
; CHECK: # %bb.0:
; CHECK-NEXT: testq %rsi, %rsi
; CHECK-NEXT: setne %cl
; CHECK-NEXT: testq %rdi, %rdi
; CHECK-NEXT: setne %al
; CHECK-NEXT: andb %cl, %al
; CHECK-NEXT: retq
%m = mul nsw nuw i64 %x, %y
%r = icmp ne i64 %m, 0
ret i1 %r
}
define <16 x i1> @mul_nuw_eq0_v16i8(<16 x i8> %x, <16 x i8> %y) {
; SSE-LABEL: mul_nuw_eq0_v16i8:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm2, %xmm2
; SSE-NEXT: pcmpeqb %xmm2, %xmm1
; SSE-NEXT: pcmpeqb %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: mul_nuw_eq0_v16i8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
; AVX-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm0
; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%m = mul nuw <16 x i8> %x, %y
%r = icmp eq <16 x i8> %m, zeroinitializer
ret <16 x i1> %r
}
define <4 x i1> @mul_nsw_ne0_v4i32(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: mul_nsw_ne0_v4i32:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm2, %xmm2
; SSE-NEXT: pcmpeqd %xmm2, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: pcmpeqd %xmm1, %xmm1
; SSE-NEXT: pxor %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: mul_nsw_ne0_v4i32:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
; AVX-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%m = mul nsw <4 x i32> %x, %y
%r = icmp ne <4 x i32> %m, zeroinitializer
ret <4 x i1> %r
}
; negative test - don't try with minsize
; TODO: SSE would be much smaller if decomposed.
define <4 x i1> @mul_nsw_ne0_v4i32_size(<4 x i32> %x, <4 x i32> %y) minsize {
; SSE-LABEL: mul_nsw_ne0_v4i32_size:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; SSE-NEXT: pmuludq %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE-NEXT: pmuludq %xmm2, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pcmpeqd %xmm0, %xmm1
; SSE-NEXT: pcmpeqd %xmm0, %xmm0
; SSE-NEXT: pxor %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: mul_nsw_ne0_v4i32_size:
; AVX: # %bb.0:
; AVX-NEXT: vpmulld %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpxor %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%m = mul nsw <4 x i32> %x, %y
%r = icmp ne <4 x i32> %m, zeroinitializer
ret <4 x i1> %r
}
|
