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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=x86_64-unknown-unknown < %s | FileCheck %s
define <4 x float> @fmul_zero_not_fast(<4 x float> %x) nounwind {
; CHECK-LABEL: fmul_zero_not_fast:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: mulps %xmm1, %xmm0
; CHECK-NEXT: retq
%r = fmul <4 x float> %x, zeroinitializer
ret <4 x float> %r
}
define <4 x float> @fmul_zero_nsz_nnan(<4 x float> %x) nounwind {
; CHECK-LABEL: fmul_zero_nsz_nnan:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%r = fmul nsz nnan <4 x float> %x, zeroinitializer
ret <4 x float> %r
}
define <4 x float> @fmul_zero_nsz_nnan_undef(<4 x float> %x) nounwind {
; CHECK-LABEL: fmul_zero_nsz_nnan_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%r = fmul nsz nnan <4 x float> %x, <float 0.0, float 0.0, float 0.0, float undef>
ret <4 x float> %r
}
define float @fmul2_f32(float %x) {
; CHECK-LABEL: fmul2_f32:
; CHECK: # %bb.0:
; CHECK-NEXT: addss %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul float %x, 2.0
ret float %y
}
; fmul 2.0, x -> fadd x, x for vectors.
define <4 x float> @fmul2_v4f32(<4 x float> %x) {
; CHECK-LABEL: fmul2_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: addps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul <4 x float> %x, <float 2.0, float 2.0, float 2.0, float 2.0>
ret <4 x float> %y
}
define <4 x float> @fmul2_v4f32_undef(<4 x float> %x) {
; CHECK-LABEL: fmul2_v4f32_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: addps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul <4 x float> %x, <float undef, float 2.0, float 2.0, float 2.0>
ret <4 x float> %y
}
define <4 x float> @constant_fold_fmul_v4f32(<4 x float> %x) {
; CHECK-LABEL: constant_fold_fmul_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps {{.*#+}} xmm0 = [8.0E+0,8.0E+0,8.0E+0,8.0E+0]
; CHECK-NEXT: retq
%y = fmul <4 x float> <float 4.0, float 4.0, float 4.0, float 4.0>, <float 2.0, float 2.0, float 2.0, float 2.0>
ret <4 x float> %y
}
define <4 x float> @constant_fold_fmul_v4f32_undef(<4 x float> %x) {
; CHECK-LABEL: constant_fold_fmul_v4f32_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps {{.*#+}} xmm0 = [8.0E+0,NaN,8.0E+0,NaN]
; CHECK-NEXT: retq
%y = fmul <4 x float> <float 4.0, float undef, float 4.0, float 4.0>, <float 2.0, float 2.0, float 2.0, float undef>
ret <4 x float> %y
}
define <4 x float> @fmul0_v4f32_nsz_nnan(<4 x float> %x) {
; CHECK-LABEL: fmul0_v4f32_nsz_nnan:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul nnan nsz <4 x float> %x, <float 0.0, float 0.0, float 0.0, float 0.0>
ret <4 x float> %y
}
define <4 x float> @fmul0_v4f32_undef(<4 x float> %x) {
; CHECK-LABEL: fmul0_v4f32_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul nnan nsz <4 x float> %x, <float undef, float 0.0, float undef, float 0.0>
ret <4 x float> %y
}
define <4 x float> @fmul_c2_c4_v4f32(<4 x float> %x) {
; CHECK-LABEL: fmul_c2_c4_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast <4 x float> %x, <float 2.0, float 2.0, float 2.0, float 2.0>
%z = fmul fast <4 x float> %y, <float 4.0, float 4.0, float 4.0, float 4.0>
ret <4 x float> %z
}
define <4 x float> @fmul_c3_c4_v4f32(<4 x float> %x) {
; CHECK-LABEL: fmul_c3_c4_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast <4 x float> %x, <float 3.0, float 3.0, float 3.0, float 3.0>
%z = fmul fast <4 x float> %y, <float 4.0, float 4.0, float 4.0, float 4.0>
ret <4 x float> %z
}
; CHECK: float 5
; CHECK: float 12
; CHECK: float 21
; CHECK: float 32
; We should be able to pre-multiply the two constant vectors.
define <4 x float> @fmul_v4f32_two_consts_no_splat(<4 x float> %x) {
; CHECK-LABEL: fmul_v4f32_two_consts_no_splat:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast <4 x float> %x, <float 1.0, float 2.0, float 3.0, float 4.0>
%z = fmul fast <4 x float> %y, <float 5.0, float 6.0, float 7.0, float 8.0>
ret <4 x float> %z
}
; Same as above, but reverse operands to make sure non-canonical form is also handled.
define <4 x float> @fmul_v4f32_two_consts_no_splat_non_canonical(<4 x float> %x) {
; CHECK-LABEL: fmul_v4f32_two_consts_no_splat_non_canonical:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, %x
%z = fmul fast <4 x float> <float 5.0, float 6.0, float 7.0, float 8.0>, %y
ret <4 x float> %z
}
; Node-level FMF and no function-level attributes.
define <4 x float> @fmul_v4f32_two_consts_no_splat_reassoc(<4 x float> %x) {
; CHECK-LABEL: fmul_v4f32_two_consts_no_splat_reassoc:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul <4 x float> %x, <float 1.0, float 2.0, float 3.0, float 4.0>
%z = fmul reassoc <4 x float> %y, <float 5.0, float 6.0, float 7.0, float 8.0>
ret <4 x float> %z
}
; Multiplication by 2.0 is a special case because that gets converted to fadd x, x.
define <4 x float> @fmul_v4f32_two_consts_no_splat_reassoc_2(<4 x float> %x) {
; CHECK-LABEL: fmul_v4f32_two_consts_no_splat_reassoc_2:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fadd <4 x float> %x, %x
%z = fmul reassoc <4 x float> %y, <float 5.0, float 6.0, float 7.0, float 8.0>
ret <4 x float> %z
}
; CHECK: float 6
; CHECK: float 14
; CHECK: float 24
; CHECK: float 36
; More than one use of a constant multiply should not inhibit the optimization.
; Instead of a chain of 2 dependent mults, this test will have 2 independent mults.
define <4 x float> @fmul_v4f32_two_consts_no_splat_multiple_use(<4 x float> %x) {
; CHECK-LABEL: fmul_v4f32_two_consts_no_splat_multiple_use:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast <4 x float> %x, <float 1.0, float 2.0, float 3.0, float 4.0>
%z = fmul fast <4 x float> %y, <float 5.0, float 6.0, float 7.0, float 8.0>
%a = fadd fast <4 x float> %y, %z
ret <4 x float> %a
}
; PR22698 - http://llvm.org/bugs/show_bug.cgi?id=22698
; Make sure that we don't infinite loop swapping constants back and forth.
; CHECK: float 24
; CHECK: float 24
; CHECK: float 24
; CHECK: float 24
define <4 x float> @PR22698_splats(<4 x float> %a) {
; CHECK-LABEL: PR22698_splats:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%mul1 = fmul fast <4 x float> <float 2.0, float 2.0, float 2.0, float 2.0>, <float 3.0, float 3.0, float 3.0, float 3.0>
%mul2 = fmul fast <4 x float> <float 4.0, float 4.0, float 4.0, float 4.0>, %mul1
%mul3 = fmul fast <4 x float> %a, %mul2
ret <4 x float> %mul3
}
; Same as above, but verify that non-splat vectors are handled correctly too.
; CHECK: float 45
; CHECK: float 120
; CHECK: float 231
; CHECK: float 384
define <4 x float> @PR22698_no_splats(<4 x float> %a) {
; CHECK-LABEL: PR22698_no_splats:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%mul1 = fmul fast <4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, <float 5.0, float 6.0, float 7.0, float 8.0>
%mul2 = fmul fast <4 x float> <float 9.0, float 10.0, float 11.0, float 12.0>, %mul1
%mul3 = fmul fast <4 x float> %a, %mul2
ret <4 x float> %mul3
}
define float @fmul_c2_c4_f32(float %x) {
; CHECK-LABEL: fmul_c2_c4_f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast float %x, 2.0
%z = fmul fast float %y, 4.0
ret float %z
}
define float @fmul_c3_c4_f32(float %x) {
; CHECK-LABEL: fmul_c3_c4_f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: retq
%y = fmul fast float %x, 3.0
%z = fmul fast float %y, 4.0
ret float %z
}
define float @fmul_fneg_fneg_f32(float %x, float %y) {
; CHECK-LABEL: fmul_fneg_fneg_f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulss %xmm1, %xmm0
; CHECK-NEXT: retq
%x.neg = fsub float -0.0, %x
%y.neg = fsub float -0.0, %y
%mul = fmul float %x.neg, %y.neg
ret float %mul
}
define <4 x float> @fmul_fneg_fneg_v4f32(<4 x float> %x, <4 x float> %y) {
; CHECK-LABEL: fmul_fneg_fneg_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps %xmm1, %xmm0
; CHECK-NEXT: retq
%x.neg = fsub <4 x float> <float -0.0, float -0.0, float -0.0, float -0.0>, %x
%y.neg = fsub <4 x float> <float -0.0, float -0.0, float -0.0, float -0.0>, %y
%mul = fmul <4 x float> %x.neg, %y.neg
ret <4 x float> %mul
}
; PR47517 - this could crash if we create 'fmul x, 0.0' nodes
; that do not constant fold in a particular order.
define float @getNegatedExpression_crash(float* %p) {
; CHECK-LABEL: getNegatedExpression_crash:
; CHECK: # %bb.0:
; CHECK-NEXT: movl $0, (%rdi)
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
store float 0.0, float* %p, align 1
%real = load float, float* %p, align 1
%r2 = fmul fast float %real, %real
%t1 = fmul fast float %real, 42.0
%t2 = fmul fast float %real, %t1
%mul_ac56 = fmul fast float %t2, %t1
%mul_ac72 = fmul fast float %r2, %mul_ac56
ret float %mul_ac72
}
|
