Mercurial > hg > CbC > CbC_llvm

comparison test/CodeGen/X86/fma-intrinsics-x86.ll @ 95:afa8332a0e37 LLVM3.8

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
LLVM 3.8
author Kaito Tokumori <e105711@ie.u-ryukyu.ac.jp>
date Tue, 13 Oct 2015 17:48:58 +0900
parents
children 7d135dc70f03
comparison
equal deleted inserted replaced
84:f3e34b893a5f 95:afa8332a0e37
1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -march=x86-64 -mcpu=corei7-avx -mattr=+fma | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -march=x86-64 -mcpu=core-avx2 -mattr=+fma,+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -march=x86-64 -mcpu=corei7-avx -mattr=+fma4 | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA4
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -mattr=+avx,-fma | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA4
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -mattr=-fma4 | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA
6
7 ; VFMADD
8 define <4 x float> @test_x86_fma_vfmadd_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
9 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_ss:
10 ; CHECK-FMA: # BB#0:
11 ; CHECK-FMA-NEXT: vfmadd213ss %xmm2, %xmm1, %xmm0
12 ; CHECK-FMA-NEXT: retq
13 ;
14 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_ss:
15 ; CHECK-FMA4: # BB#0:
16 ; CHECK-FMA4-NEXT: vfmaddss %xmm2, %xmm1, %xmm0, %xmm0
17 ; CHECK-FMA4-NEXT: retq
18 %res = call <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
19 ret <4 x float> %res
20 }
21 declare <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>)
22
23 define <2 x double> @test_x86_fma_vfmadd_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
24 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_sd:
25 ; CHECK-FMA: # BB#0:
26 ; CHECK-FMA-NEXT: vfmadd213sd %xmm2, %xmm1, %xmm0
27 ; CHECK-FMA-NEXT: retq
28 ;
29 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_sd:
30 ; CHECK-FMA4: # BB#0:
31 ; CHECK-FMA4-NEXT: vfmaddsd %xmm2, %xmm1, %xmm0, %xmm0
32 ; CHECK-FMA4-NEXT: retq
33 %res = call <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
34 ret <2 x double> %res
35 }
36 declare <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double>, <2 x double>, <2 x double>)
37
38 define <4 x float> @test_x86_fma_vfmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
39 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_ps:
40 ; CHECK-FMA: # BB#0:
41 ; CHECK-FMA-NEXT: vfmadd213ps %xmm2, %xmm1, %xmm0
42 ; CHECK-FMA-NEXT: retq
43 ;
44 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_ps:
45 ; CHECK-FMA4: # BB#0:
46 ; CHECK-FMA4-NEXT: vfmaddps %xmm2, %xmm1, %xmm0, %xmm0
47 ; CHECK-FMA4-NEXT: retq
48 %res = call <4 x float> @llvm.x86.fma.vfmadd.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
49 ret <4 x float> %res
50 }
51 declare <4 x float> @llvm.x86.fma.vfmadd.ps(<4 x float>, <4 x float>, <4 x float>)
52
53 define <2 x double> @test_x86_fma_vfmadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
54 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_pd:
55 ; CHECK-FMA: # BB#0:
56 ; CHECK-FMA-NEXT: vfmadd213pd %xmm2, %xmm1, %xmm0
57 ; CHECK-FMA-NEXT: retq
58 ;
59 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_pd:
60 ; CHECK-FMA4: # BB#0:
61 ; CHECK-FMA4-NEXT: vfmaddpd %xmm2, %xmm1, %xmm0, %xmm0
62 ; CHECK-FMA4-NEXT: retq
63 %res = call <2 x double> @llvm.x86.fma.vfmadd.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
64 ret <2 x double> %res
65 }
66 declare <2 x double> @llvm.x86.fma.vfmadd.pd(<2 x double>, <2 x double>, <2 x double>)
67
68 define <8 x float> @test_x86_fma_vfmadd_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
69 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_ps_256:
70 ; CHECK-FMA: # BB#0:
71 ; CHECK-FMA-NEXT: vfmadd213ps %ymm2, %ymm1, %ymm0
72 ; CHECK-FMA-NEXT: retq
73 ;
74 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_ps_256:
75 ; CHECK-FMA4: # BB#0:
76 ; CHECK-FMA4-NEXT: vfmaddps %ymm2, %ymm1, %ymm0, %ymm0
77 ; CHECK-FMA4-NEXT: retq
78 %res = call <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
79 ret <8 x float> %res
80 }
81 declare <8 x float> @llvm.x86.fma.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>)
82
83 define <4 x double> @test_x86_fma_vfmadd_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
84 ; CHECK-FMA-LABEL: test_x86_fma_vfmadd_pd_256:
85 ; CHECK-FMA: # BB#0:
86 ; CHECK-FMA-NEXT: vfmadd213pd %ymm2, %ymm1, %ymm0
87 ; CHECK-FMA-NEXT: retq
88 ;
89 ; CHECK-FMA4-LABEL: test_x86_fma_vfmadd_pd_256:
90 ; CHECK-FMA4: # BB#0:
91 ; CHECK-FMA4-NEXT: vfmaddpd %ymm2, %ymm1, %ymm0, %ymm0
92 ; CHECK-FMA4-NEXT: retq
93 %res = call <4 x double> @llvm.x86.fma.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
94 ret <4 x double> %res
95 }
96 declare <4 x double> @llvm.x86.fma.vfmadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
97
98 ; VFMSUB
99 define <4 x float> @test_x86_fma_vfmsub_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
100 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_ss:
101 ; CHECK-FMA: # BB#0:
102 ; CHECK-FMA-NEXT: vfmsub213ss %xmm2, %xmm1, %xmm0
103 ; CHECK-FMA-NEXT: retq
104 ;
105 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_ss:
106 ; CHECK-FMA4: # BB#0:
107 ; CHECK-FMA4-NEXT: vfmsubss %xmm2, %xmm1, %xmm0, %xmm0
108 ; CHECK-FMA4-NEXT: retq
109 %res = call <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
110 ret <4 x float> %res
111 }
112 declare <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float>, <4 x float>, <4 x float>)
113
114 define <2 x double> @test_x86_fma_vfmsub_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
115 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_sd:
116 ; CHECK-FMA: # BB#0:
117 ; CHECK-FMA-NEXT: vfmsub213sd %xmm2, %xmm1, %xmm0
118 ; CHECK-FMA-NEXT: retq
119 ;
120 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_sd:
121 ; CHECK-FMA4: # BB#0:
122 ; CHECK-FMA4-NEXT: vfmsubsd %xmm2, %xmm1, %xmm0, %xmm0
123 ; CHECK-FMA4-NEXT: retq
124 %res = call <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
125 ret <2 x double> %res
126 }
127 declare <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double>, <2 x double>, <2 x double>)
128
129 define <4 x float> @test_x86_fma_vfmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
130 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_ps:
131 ; CHECK-FMA: # BB#0:
132 ; CHECK-FMA-NEXT: vfmsub213ps %xmm2, %xmm1, %xmm0
133 ; CHECK-FMA-NEXT: retq
134 ;
135 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_ps:
136 ; CHECK-FMA4: # BB#0:
137 ; CHECK-FMA4-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0
138 ; CHECK-FMA4-NEXT: retq
139 %res = call <4 x float> @llvm.x86.fma.vfmsub.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
140 ret <4 x float> %res
141 }
142 declare <4 x float> @llvm.x86.fma.vfmsub.ps(<4 x float>, <4 x float>, <4 x float>)
143
144 define <2 x double> @test_x86_fma_vfmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
145 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_pd:
146 ; CHECK-FMA: # BB#0:
147 ; CHECK-FMA-NEXT: vfmsub213pd %xmm2, %xmm1, %xmm0
148 ; CHECK-FMA-NEXT: retq
149 ;
150 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_pd:
151 ; CHECK-FMA4: # BB#0:
152 ; CHECK-FMA4-NEXT: vfmsubpd %xmm2, %xmm1, %xmm0, %xmm0
153 ; CHECK-FMA4-NEXT: retq
154 %res = call <2 x double> @llvm.x86.fma.vfmsub.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
155 ret <2 x double> %res
156 }
157 declare <2 x double> @llvm.x86.fma.vfmsub.pd(<2 x double>, <2 x double>, <2 x double>)
158
159 define <8 x float> @test_x86_fma_vfmsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
160 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_ps_256:
161 ; CHECK-FMA: # BB#0:
162 ; CHECK-FMA-NEXT: vfmsub213ps %ymm2, %ymm1, %ymm0
163 ; CHECK-FMA-NEXT: retq
164 ;
165 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_ps_256:
166 ; CHECK-FMA4: # BB#0:
167 ; CHECK-FMA4-NEXT: vfmsubps %ymm2, %ymm1, %ymm0, %ymm0
168 ; CHECK-FMA4-NEXT: retq
169 %res = call <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
170 ret <8 x float> %res
171 }
172 declare <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float>, <8 x float>, <8 x float>)
173
174 define <4 x double> @test_x86_fma_vfmsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
175 ; CHECK-FMA-LABEL: test_x86_fma_vfmsub_pd_256:
176 ; CHECK-FMA: # BB#0:
177 ; CHECK-FMA-NEXT: vfmsub213pd %ymm2, %ymm1, %ymm0
178 ; CHECK-FMA-NEXT: retq
179 ;
180 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsub_pd_256:
181 ; CHECK-FMA4: # BB#0:
182 ; CHECK-FMA4-NEXT: vfmsubpd %ymm2, %ymm1, %ymm0, %ymm0
183 ; CHECK-FMA4-NEXT: retq
184 %res = call <4 x double> @llvm.x86.fma.vfmsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
185 ret <4 x double> %res
186 }
187 declare <4 x double> @llvm.x86.fma.vfmsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
188
189 ; VFNMADD
190 define <4 x float> @test_x86_fma_vfnmadd_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
191 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_ss:
192 ; CHECK-FMA: # BB#0:
193 ; CHECK-FMA-NEXT: vfnmadd213ss %xmm2, %xmm1, %xmm0
194 ; CHECK-FMA-NEXT: retq
195 ;
196 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_ss:
197 ; CHECK-FMA4: # BB#0:
198 ; CHECK-FMA4-NEXT: vfnmaddss %xmm2, %xmm1, %xmm0, %xmm0
199 ; CHECK-FMA4-NEXT: retq
200 %res = call <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
201 ret <4 x float> %res
202 }
203 declare <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float>, <4 x float>, <4 x float>)
204
205 define <2 x double> @test_x86_fma_vfnmadd_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
206 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_sd:
207 ; CHECK-FMA: # BB#0:
208 ; CHECK-FMA-NEXT: vfnmadd213sd %xmm2, %xmm1, %xmm0
209 ; CHECK-FMA-NEXT: retq
210 ;
211 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_sd:
212 ; CHECK-FMA4: # BB#0:
213 ; CHECK-FMA4-NEXT: vfnmaddsd %xmm2, %xmm1, %xmm0, %xmm0
214 ; CHECK-FMA4-NEXT: retq
215 %res = call <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
216 ret <2 x double> %res
217 }
218 declare <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double>, <2 x double>, <2 x double>)
219
220 define <4 x float> @test_x86_fma_vfnmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
221 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_ps:
222 ; CHECK-FMA: # BB#0:
223 ; CHECK-FMA-NEXT: vfnmadd213ps %xmm2, %xmm1, %xmm0
224 ; CHECK-FMA-NEXT: retq
225 ;
226 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_ps:
227 ; CHECK-FMA4: # BB#0:
228 ; CHECK-FMA4-NEXT: vfnmaddps %xmm2, %xmm1, %xmm0, %xmm0
229 ; CHECK-FMA4-NEXT: retq
230 %res = call <4 x float> @llvm.x86.fma.vfnmadd.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
231 ret <4 x float> %res
232 }
233 declare <4 x float> @llvm.x86.fma.vfnmadd.ps(<4 x float>, <4 x float>, <4 x float>)
234
235 define <2 x double> @test_x86_fma_vfnmadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
236 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_pd:
237 ; CHECK-FMA: # BB#0:
238 ; CHECK-FMA-NEXT: vfnmadd213pd %xmm2, %xmm1, %xmm0
239 ; CHECK-FMA-NEXT: retq
240 ;
241 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_pd:
242 ; CHECK-FMA4: # BB#0:
243 ; CHECK-FMA4-NEXT: vfnmaddpd %xmm2, %xmm1, %xmm0, %xmm0
244 ; CHECK-FMA4-NEXT: retq
245 %res = call <2 x double> @llvm.x86.fma.vfnmadd.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
246 ret <2 x double> %res
247 }
248 declare <2 x double> @llvm.x86.fma.vfnmadd.pd(<2 x double>, <2 x double>, <2 x double>)
249
250 define <8 x float> @test_x86_fma_vfnmadd_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
251 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_ps_256:
252 ; CHECK-FMA: # BB#0:
253 ; CHECK-FMA-NEXT: vfnmadd213ps %ymm2, %ymm1, %ymm0
254 ; CHECK-FMA-NEXT: retq
255 ;
256 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_ps_256:
257 ; CHECK-FMA4: # BB#0:
258 ; CHECK-FMA4-NEXT: vfnmaddps %ymm2, %ymm1, %ymm0, %ymm0
259 ; CHECK-FMA4-NEXT: retq
260 %res = call <8 x float> @llvm.x86.fma.vfnmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
261 ret <8 x float> %res
262 }
263 declare <8 x float> @llvm.x86.fma.vfnmadd.ps.256(<8 x float>, <8 x float>, <8 x float>)
264
265 define <4 x double> @test_x86_fma_vfnmadd_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
266 ; CHECK-FMA-LABEL: test_x86_fma_vfnmadd_pd_256:
267 ; CHECK-FMA: # BB#0:
268 ; CHECK-FMA-NEXT: vfnmadd213pd %ymm2, %ymm1, %ymm0
269 ; CHECK-FMA-NEXT: retq
270 ;
271 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmadd_pd_256:
272 ; CHECK-FMA4: # BB#0:
273 ; CHECK-FMA4-NEXT: vfnmaddpd %ymm2, %ymm1, %ymm0, %ymm0
274 ; CHECK-FMA4-NEXT: retq
275 %res = call <4 x double> @llvm.x86.fma.vfnmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
276 ret <4 x double> %res
277 }
278 declare <4 x double> @llvm.x86.fma.vfnmadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
279
280 ; VFNMSUB
281 define <4 x float> @test_x86_fma_vfnmsub_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
282 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_ss:
283 ; CHECK-FMA: # BB#0:
284 ; CHECK-FMA-NEXT: vfnmsub213ss %xmm2, %xmm1, %xmm0
285 ; CHECK-FMA-NEXT: retq
286 ;
287 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_ss:
288 ; CHECK-FMA4: # BB#0:
289 ; CHECK-FMA4-NEXT: vfnmsubss %xmm2, %xmm1, %xmm0, %xmm0
290 ; CHECK-FMA4-NEXT: retq
291 %res = call <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
292 ret <4 x float> %res
293 }
294 declare <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float>, <4 x float>, <4 x float>)
295
296 define <2 x double> @test_x86_fma_vfnmsub_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
297 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_sd:
298 ; CHECK-FMA: # BB#0:
299 ; CHECK-FMA-NEXT: vfnmsub213sd %xmm2, %xmm1, %xmm0
300 ; CHECK-FMA-NEXT: retq
301 ;
302 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_sd:
303 ; CHECK-FMA4: # BB#0:
304 ; CHECK-FMA4-NEXT: vfnmsubsd %xmm2, %xmm1, %xmm0, %xmm0
305 ; CHECK-FMA4-NEXT: retq
306 %res = call <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
307 ret <2 x double> %res
308 }
309 declare <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double>, <2 x double>, <2 x double>)
310
311 define <4 x float> @test_x86_fma_vfnmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
312 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_ps:
313 ; CHECK-FMA: # BB#0:
314 ; CHECK-FMA-NEXT: vfnmsub213ps %xmm2, %xmm1, %xmm0
315 ; CHECK-FMA-NEXT: retq
316 ;
317 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_ps:
318 ; CHECK-FMA4: # BB#0:
319 ; CHECK-FMA4-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0
320 ; CHECK-FMA4-NEXT: retq
321 %res = call <4 x float> @llvm.x86.fma.vfnmsub.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
322 ret <4 x float> %res
323 }
324 declare <4 x float> @llvm.x86.fma.vfnmsub.ps(<4 x float>, <4 x float>, <4 x float>)
325
326 define <2 x double> @test_x86_fma_vfnmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
327 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_pd:
328 ; CHECK-FMA: # BB#0:
329 ; CHECK-FMA-NEXT: vfnmsub213pd %xmm2, %xmm1, %xmm0
330 ; CHECK-FMA-NEXT: retq
331 ;
332 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_pd:
333 ; CHECK-FMA4: # BB#0:
334 ; CHECK-FMA4-NEXT: vfnmsubpd %xmm2, %xmm1, %xmm0, %xmm0
335 ; CHECK-FMA4-NEXT: retq
336 %res = call <2 x double> @llvm.x86.fma.vfnmsub.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
337 ret <2 x double> %res
338 }
339 declare <2 x double> @llvm.x86.fma.vfnmsub.pd(<2 x double>, <2 x double>, <2 x double>)
340
341 define <8 x float> @test_x86_fma_vfnmsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
342 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_ps_256:
343 ; CHECK-FMA: # BB#0:
344 ; CHECK-FMA-NEXT: vfnmsub213ps %ymm2, %ymm1, %ymm0
345 ; CHECK-FMA-NEXT: retq
346 ;
347 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_ps_256:
348 ; CHECK-FMA4: # BB#0:
349 ; CHECK-FMA4-NEXT: vfnmsubps %ymm2, %ymm1, %ymm0, %ymm0
350 ; CHECK-FMA4-NEXT: retq
351 %res = call <8 x float> @llvm.x86.fma.vfnmsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
352 ret <8 x float> %res
353 }
354 declare <8 x float> @llvm.x86.fma.vfnmsub.ps.256(<8 x float>, <8 x float>, <8 x float>)
355
356 define <4 x double> @test_x86_fma_vfnmsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
357 ; CHECK-FMA-LABEL: test_x86_fma_vfnmsub_pd_256:
358 ; CHECK-FMA: # BB#0:
359 ; CHECK-FMA-NEXT: vfnmsub213pd %ymm2, %ymm1, %ymm0
360 ; CHECK-FMA-NEXT: retq
361 ;
362 ; CHECK-FMA4-LABEL: test_x86_fma_vfnmsub_pd_256:
363 ; CHECK-FMA4: # BB#0:
364 ; CHECK-FMA4-NEXT: vfnmsubpd %ymm2, %ymm1, %ymm0, %ymm0
365 ; CHECK-FMA4-NEXT: retq
366 %res = call <4 x double> @llvm.x86.fma.vfnmsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
367 ret <4 x double> %res
368 }
369 declare <4 x double> @llvm.x86.fma.vfnmsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
370
371 ; VFMADDSUB
372 define <4 x float> @test_x86_fma_vfmaddsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
373 ; CHECK-FMA-LABEL: test_x86_fma_vfmaddsub_ps:
374 ; CHECK-FMA: # BB#0:
375 ; CHECK-FMA-NEXT: vfmaddsub213ps %xmm2, %xmm1, %xmm0
376 ; CHECK-FMA-NEXT: retq
377 ;
378 ; CHECK-FMA4-LABEL: test_x86_fma_vfmaddsub_ps:
379 ; CHECK-FMA4: # BB#0:
380 ; CHECK-FMA4-NEXT: vfmaddsubps %xmm2, %xmm1, %xmm0, %xmm0
381 ; CHECK-FMA4-NEXT: retq
382 %res = call <4 x float> @llvm.x86.fma.vfmaddsub.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
383 ret <4 x float> %res
384 }
385 declare <4 x float> @llvm.x86.fma.vfmaddsub.ps(<4 x float>, <4 x float>, <4 x float>)
386
387 define <2 x double> @test_x86_fma_vfmaddsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
388 ; CHECK-FMA-LABEL: test_x86_fma_vfmaddsub_pd:
389 ; CHECK-FMA: # BB#0:
390 ; CHECK-FMA-NEXT: vfmaddsub213pd %xmm2, %xmm1, %xmm0
391 ; CHECK-FMA-NEXT: retq
392 ;
393 ; CHECK-FMA4-LABEL: test_x86_fma_vfmaddsub_pd:
394 ; CHECK-FMA4: # BB#0:
395 ; CHECK-FMA4-NEXT: vfmaddsubpd %xmm2, %xmm1, %xmm0, %xmm0
396 ; CHECK-FMA4-NEXT: retq
397 %res = call <2 x double> @llvm.x86.fma.vfmaddsub.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
398 ret <2 x double> %res
399 }
400 declare <2 x double> @llvm.x86.fma.vfmaddsub.pd(<2 x double>, <2 x double>, <2 x double>)
401
402 define <8 x float> @test_x86_fma_vfmaddsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
403 ; CHECK-FMA-LABEL: test_x86_fma_vfmaddsub_ps_256:
404 ; CHECK-FMA: # BB#0:
405 ; CHECK-FMA-NEXT: vfmaddsub213ps %ymm2, %ymm1, %ymm0
406 ; CHECK-FMA-NEXT: retq
407 ;
408 ; CHECK-FMA4-LABEL: test_x86_fma_vfmaddsub_ps_256:
409 ; CHECK-FMA4: # BB#0:
410 ; CHECK-FMA4-NEXT: vfmaddsubps %ymm2, %ymm1, %ymm0, %ymm0
411 ; CHECK-FMA4-NEXT: retq
412 %res = call <8 x float> @llvm.x86.fma.vfmaddsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
413 ret <8 x float> %res
414 }
415 declare <8 x float> @llvm.x86.fma.vfmaddsub.ps.256(<8 x float>, <8 x float>, <8 x float>)
416
417 define <4 x double> @test_x86_fma_vfmaddsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
418 ; CHECK-FMA-LABEL: test_x86_fma_vfmaddsub_pd_256:
419 ; CHECK-FMA: # BB#0:
420 ; CHECK-FMA-NEXT: vfmaddsub213pd %ymm2, %ymm1, %ymm0
421 ; CHECK-FMA-NEXT: retq
422 ;
423 ; CHECK-FMA4-LABEL: test_x86_fma_vfmaddsub_pd_256:
424 ; CHECK-FMA4: # BB#0:
425 ; CHECK-FMA4-NEXT: vfmaddsubpd %ymm2, %ymm1, %ymm0, %ymm0
426 ; CHECK-FMA4-NEXT: retq
427 %res = call <4 x double> @llvm.x86.fma.vfmaddsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
428 ret <4 x double> %res
429 }
430 declare <4 x double> @llvm.x86.fma.vfmaddsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
431
432 ; VFMSUBADD
433 define <4 x float> @test_x86_fma_vfmsubadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
434 ; CHECK-FMA-LABEL: test_x86_fma_vfmsubadd_ps:
435 ; CHECK-FMA: # BB#0:
436 ; CHECK-FMA-NEXT: vfmsubadd213ps %xmm2, %xmm1, %xmm0
437 ; CHECK-FMA-NEXT: retq
438 ;
439 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsubadd_ps:
440 ; CHECK-FMA4: # BB#0:
441 ; CHECK-FMA4-NEXT: vfmsubaddps %xmm2, %xmm1, %xmm0, %xmm0
442 ; CHECK-FMA4-NEXT: retq
443 %res = call <4 x float> @llvm.x86.fma.vfmsubadd.ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
444 ret <4 x float> %res
445 }
446 declare <4 x float> @llvm.x86.fma.vfmsubadd.ps(<4 x float>, <4 x float>, <4 x float>)
447
448 define <2 x double> @test_x86_fma_vfmsubadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
449 ; CHECK-FMA-LABEL: test_x86_fma_vfmsubadd_pd:
450 ; CHECK-FMA: # BB#0:
451 ; CHECK-FMA-NEXT: vfmsubadd213pd %xmm2, %xmm1, %xmm0
452 ; CHECK-FMA-NEXT: retq
453 ;
454 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsubadd_pd:
455 ; CHECK-FMA4: # BB#0:
456 ; CHECK-FMA4-NEXT: vfmsubaddpd %xmm2, %xmm1, %xmm0, %xmm0
457 ; CHECK-FMA4-NEXT: retq
458 %res = call <2 x double> @llvm.x86.fma.vfmsubadd.pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
459 ret <2 x double> %res
460 }
461 declare <2 x double> @llvm.x86.fma.vfmsubadd.pd(<2 x double>, <2 x double>, <2 x double>)
462
463 define <8 x float> @test_x86_fma_vfmsubadd_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
464 ; CHECK-FMA-LABEL: test_x86_fma_vfmsubadd_ps_256:
465 ; CHECK-FMA: # BB#0:
466 ; CHECK-FMA-NEXT: vfmsubadd213ps %ymm2, %ymm1, %ymm0
467 ; CHECK-FMA-NEXT: retq
468 ;
469 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsubadd_ps_256:
470 ; CHECK-FMA4: # BB#0:
471 ; CHECK-FMA4-NEXT: vfmsubaddps %ymm2, %ymm1, %ymm0, %ymm0
472 ; CHECK-FMA4-NEXT: retq
473 %res = call <8 x float> @llvm.x86.fma.vfmsubadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
474 ret <8 x float> %res
475 }
476 declare <8 x float> @llvm.x86.fma.vfmsubadd.ps.256(<8 x float>, <8 x float>, <8 x float>)
477
478 define <4 x double> @test_x86_fma_vfmsubadd_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
479 ; CHECK-FMA-LABEL: test_x86_fma_vfmsubadd_pd_256:
480 ; CHECK-FMA: # BB#0:
481 ; CHECK-FMA-NEXT: vfmsubadd213pd %ymm2, %ymm1, %ymm0
482 ; CHECK-FMA-NEXT: retq
483 ;
484 ; CHECK-FMA4-LABEL: test_x86_fma_vfmsubadd_pd_256:
485 ; CHECK-FMA4: # BB#0:
486 ; CHECK-FMA4-NEXT: vfmsubaddpd %ymm2, %ymm1, %ymm0, %ymm0
487 ; CHECK-FMA4-NEXT: retq
488 %res = call <4 x double> @llvm.x86.fma.vfmsubadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
489 ret <4 x double> %res
490 }
491 declare <4 x double> @llvm.x86.fma.vfmsubadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
492
493 attributes #0 = { nounwind }