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
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=sse2 < %s | FileCheck %s
; PR22428: https://llvm.org/bugs/show_bug.cgi?id=22428
; f1, f2, f3, and f4 should use an integer logic instruction.
; f5, f6, f9, and f10 should use an FP (SSE) logic instruction.
;
; f7 and f8 are less clear.
;
; For f7 and f8, the SSE instructions don't take immediate operands, so if we
; use one of those, we either have to load a constant from memory or move the
; scalar immediate value from an integer register over to an SSE register.
; Optimizing for size may affect that decision. Also, note that there are no
; scalar versions of the FP logic ops, so if we want to fold a load into a
; logic op, we have to load or splat a 16-byte vector constant.
; 1 FP operand, 1 int operand, int result
define i32 @f1(float %x, i32 %y) {
; CHECK-LABEL: f1:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %xmm0, %eax
; CHECK-NEXT: andl %edi, %eax
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %bc1, %y
ret i32 %and
}
; Swap operands of the logic op.
define i32 @f2(float %x, i32 %y) {
; CHECK-LABEL: f2:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %xmm0, %eax
; CHECK-NEXT: andl %edi, %eax
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %y, %bc1
ret i32 %and
}
; 1 FP operand, 1 constant operand, int result
define i32 @f3(float %x) {
; CHECK-LABEL: f3:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %xmm0, %eax
; CHECK-NEXT: andl $1, %eax
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %bc1, 1
ret i32 %and
}
; Swap operands of the logic op.
define i32 @f4(float %x) {
; CHECK-LABEL: f4:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %xmm0, %eax
; CHECK-NEXT: andl $2, %eax
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 2, %bc1
ret i32 %and
}
; 1 FP operand, 1 integer operand, FP result
define float @f5(float %x, i32 %y) {
; CHECK-LABEL: f5:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %edi, %xmm1
; CHECK-NEXT: pand %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %bc1, %y
%bc2 = bitcast i32 %and to float
ret float %bc2
}
; Swap operands of the logic op.
define float @f6(float %x, i32 %y) {
; CHECK-LABEL: f6:
; CHECK: # %bb.0:
; CHECK-NEXT: movd %edi, %xmm1
; CHECK-NEXT: pand %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %y, %bc1
%bc2 = bitcast i32 %and to float
ret float %bc2
}
; 1 FP operand, 1 constant operand, FP result
define float @f7(float %x) {
; CHECK-LABEL: f7:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %bc1, 3
%bc2 = bitcast i32 %and to float
ret float %bc2
}
; Swap operands of the logic op.
define float @f8(float %x) {
; CHECK-LABEL: f8:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 4, %bc1
%bc2 = bitcast i32 %and to float
ret float %bc2
}
; 2 FP operands, int result
define i32 @f9(float %x, float %y) {
; CHECK-LABEL: f9:
; CHECK: # %bb.0:
; CHECK-NEXT: pand %xmm1, %xmm0
; CHECK-NEXT: movd %xmm0, %eax
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%bc2 = bitcast float %y to i32
%and = and i32 %bc1, %bc2
ret i32 %and
}
; 2 FP operands, FP result
define float @f10(float %x, float %y) {
; CHECK-LABEL: f10:
; CHECK: # %bb.0:
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%bc2 = bitcast float %y to i32
%and = and i32 %bc1, %bc2
%bc3 = bitcast i32 %and to float
ret float %bc3
}
define float @or(float %x, float %y) {
; CHECK-LABEL: or:
; CHECK: # %bb.0:
; CHECK-NEXT: orps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%bc2 = bitcast float %y to i32
%and = or i32 %bc1, %bc2
%bc3 = bitcast i32 %and to float
ret float %bc3
}
define float @xor(float %x, float %y) {
; CHECK-LABEL: xor:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%bc2 = bitcast float %y to i32
%and = xor i32 %bc1, %bc2
%bc3 = bitcast i32 %and to float
ret float %bc3
}
define float @f7_or(float %x) {
; CHECK-LABEL: f7_or:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: orps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = or i32 %bc1, 3
%bc2 = bitcast i32 %and to float
ret float %bc2
}
define float @f7_xor(float %x) {
; CHECK-LABEL: f7_xor:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: xorps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = xor i32 %bc1, 3
%bc2 = bitcast i32 %and to float
ret float %bc2
}
; Make sure that doubles work too.
define double @doubles(double %x, double %y) {
; CHECK-LABEL: doubles:
; CHECK: # %bb.0:
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast double %x to i64
%bc2 = bitcast double %y to i64
%and = and i64 %bc1, %bc2
%bc3 = bitcast i64 %and to double
ret double %bc3
}
define double @f7_double(double %x) {
; CHECK-LABEL: f7_double:
; CHECK: # %bb.0:
; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast double %x to i64
%and = and i64 %bc1, 3
%bc2 = bitcast i64 %and to double
ret double %bc2
}
; Grabbing the sign bit is a special case that could be handled
; by movmskps/movmskpd, but if we're not shifting it over, then
; a simple FP logic op is cheaper.
define float @movmsk(float %x) {
; CHECK-LABEL: movmsk:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: andps %xmm1, %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%and = and i32 %bc1, 2147483648
%bc2 = bitcast i32 %and to float
ret float %bc2
}
define double @bitcast_fabs(double %x) {
; CHECK-LABEL: bitcast_fabs:
; CHECK: # %bb.0:
; CHECK-NEXT: andps {{.*}}(%rip), %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast double %x to i64
%and = and i64 %bc1, 9223372036854775807
%bc2 = bitcast i64 %and to double
ret double %bc2
}
define float @bitcast_fneg(float %x) {
; CHECK-LABEL: bitcast_fneg:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps {{.*}}(%rip), %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast float %x to i32
%xor = xor i32 %bc1, 2147483648
%bc2 = bitcast i32 %xor to float
ret float %bc2
}
define <2 x double> @bitcast_fabs_vec(<2 x double> %x) {
; CHECK-LABEL: bitcast_fabs_vec:
; CHECK: # %bb.0:
; CHECK-NEXT: andps {{.*}}(%rip), %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast <2 x double> %x to <2 x i64>
%and = and <2 x i64> %bc1, <i64 9223372036854775807, i64 9223372036854775807>
%bc2 = bitcast <2 x i64> %and to <2 x double>
ret <2 x double> %bc2
}
define <4 x float> @bitcast_fneg_vec(<4 x float> %x) {
; CHECK-LABEL: bitcast_fneg_vec:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps {{.*}}(%rip), %xmm0
; CHECK-NEXT: retq
%bc1 = bitcast <4 x float> %x to <4 x i32>
%xor = xor <4 x i32> %bc1, <i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648>
%bc2 = bitcast <4 x i32> %xor to <4 x float>
ret <4 x float> %bc2
}
|
