Debian Codegraph @

gcc/testsuite/gcc.target/aarch64/sve2/acle/asm/mls_bf16.c

raw
1 /* { dg-do assemble { target aarch64_asm_sve-b16b16_ok } } */
2 /* { dg-do compile { target { ! aarch64_asm_sve-b16b16_ok } } } */
3 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
4
5 #include "test_sve_acle.h"
6
7 #pragma GCC target "+sve-b16b16"
8 #ifdef STREAMING_COMPATIBLE
9 #pragma GCC target "+sme2"
10 #endif
11
12 /*
13 ** mls_bf16_m_tied1:
14 ** bfmls z0\.h, p0/m, z1\.h, z2\.h
15 ** ret
16 */
17 TEST_UNIFORM_Z (mls_bf16_m_tied1, svbfloat16_t,
18 z0 = svmls_bf16_m (p0, z0, z1, z2),
19 z0 = svmls_m (p0, z0, z1, z2))
20
21 /*
22 ** mls_bf16_m_tied2:
23 ** mov (z[0-9]+)\.d, z0\.d
24 ** movprfx z0, z1
25 ** bfmls z0\.h, p0/m, \1\.h, z2\.h
26 ** ret
27 */
28 TEST_UNIFORM_Z (mls_bf16_m_tied2, svbfloat16_t,
29 z0 = svmls_bf16_m (p0, z1, z0, z2),
30 z0 = svmls_m (p0, z1, z0, z2))
31
32 /*
33 ** mls_bf16_m_tied3:
34 ** mov (z[0-9]+)\.d, z0\.d
35 ** movprfx z0, z1
36 ** bfmls z0\.h, p0/m, z2\.h, \1\.h
37 ** ret
38 */
39 TEST_UNIFORM_Z (mls_bf16_m_tied3, svbfloat16_t,
40 z0 = svmls_bf16_m (p0, z1, z2, z0),
41 z0 = svmls_m (p0, z1, z2, z0))
42
43 /*
44 ** mls_bf16_m_untied:
45 ** movprfx z0, z1
46 ** bfmls z0\.h, p0/m, z2\.h, z3\.h
47 ** ret
48 */
49 TEST_UNIFORM_Z (mls_bf16_m_untied, svbfloat16_t,
50 z0 = svmls_bf16_m (p0, z1, z2, z3),
51 z0 = svmls_m (p0, z1, z2, z3))
52
53 /*
54 ** mls_h4_bf16_m_tied1:
55 ** mov (z[0-9]+\.h), h4
56 ** bfmls z0\.h, p0/m, z1\.h, \1
57 ** ret
58 */
59 TEST_UNIFORM_ZD (mls_h4_bf16_m_tied1, svbfloat16_t, __bf16,
60 z0 = svmls_n_bf16_m (p0, z0, z1, d4),
61 z0 = svmls_m (p0, z0, z1, d4))
62
63 /*
64 ** mls_h4_bf16_m_untied:
65 ** mov (z[0-9]+\.h), h4
66 ** movprfx z0, z1
67 ** bfmls z0\.h, p0/m, z2\.h, \1
68 ** ret
69 */
70 TEST_UNIFORM_ZD (mls_h4_bf16_m_untied, svbfloat16_t, __bf16,
71 z0 = svmls_n_bf16_m (p0, z1, z2, d4),
72 z0 = svmls_m (p0, z1, z2, d4))
73
74 /*
75 ** mls_2_bf16_m_tied1:
76 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
77 ** bfmls z0\.h, p0/m, z1\.h, \1
78 ** ret
79 */
80 TEST_UNIFORM_Z (mls_2_bf16_m_tied1, svbfloat16_t,
81 z0 = svmls_n_bf16_m (p0, z0, z1, 2),
82 z0 = svmls_m (p0, z0, z1, 2))
83
84 /*
85 ** mls_2_bf16_m_untied:
86 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
87 ** movprfx z0, z1
88 ** bfmls z0\.h, p0/m, z2\.h, \1
89 ** ret
90 */
91 TEST_UNIFORM_Z (mls_2_bf16_m_untied, svbfloat16_t,
92 z0 = svmls_n_bf16_m (p0, z1, z2, 2),
93 z0 = svmls_m (p0, z1, z2, 2))
94
95 /*
96 ** mls_bf16_z_tied1:
97 ** movprfx z0\.h, p0/z, z0\.h
98 ** bfmls z0\.h, p0/m, z1\.h, z2\.h
99 ** ret
100 */
101 TEST_UNIFORM_Z (mls_bf16_z_tied1, svbfloat16_t,
102 z0 = svmls_bf16_z (p0, z0, z1, z2),
103 z0 = svmls_z (p0, z0, z1, z2))
104
105 /*
106 ** mls_bf16_z_tied2:
107 ** mov (z[0-9]+)\.d, z0\.d
108 ** movprfx z0\.h, p0/z, z1\.h
109 ** bfmls z0\.h, p0/m, \1\.h, z2\.h
110 ** ret
111 */
112 TEST_UNIFORM_Z (mls_bf16_z_tied2, svbfloat16_t,
113 z0 = svmls_bf16_z (p0, z1, z0, z2),
114 z0 = svmls_z (p0, z1, z0, z2))
115
116 /*
117 ** mls_bf16_z_tied3:
118 ** mov (z[0-9]+)\.d, z0\.d
119 ** movprfx z0\.h, p0/z, z1\.h
120 ** bfmls z0\.h, p0/m, z2\.h, \1\.h
121 ** ret
122 */
123 TEST_UNIFORM_Z (mls_bf16_z_tied3, svbfloat16_t,
124 z0 = svmls_bf16_z (p0, z1, z2, z0),
125 z0 = svmls_z (p0, z1, z2, z0))
126
127 /*
128 ** mls_bf16_z_untied:
129 ** movprfx z0\.h, p0/z, z1\.h
130 ** bfmls z0\.h, p0/m, z2\.h, z3\.h
131 ** ret
132 */
133 TEST_UNIFORM_Z (mls_bf16_z_untied, svbfloat16_t,
134 z0 = svmls_bf16_z (p0, z1, z2, z3),
135 z0 = svmls_z (p0, z1, z2, z3))
136
137 /*
138 ** mls_h4_bf16_z_tied1:
139 ** mov (z[0-9]+\.h), h4
140 ** movprfx z0\.h, p0/z, z0\.h
141 ** bfmls z0\.h, p0/m, z1\.h, \1
142 ** ret
143 */
144 TEST_UNIFORM_ZD (mls_h4_bf16_z_tied1, svbfloat16_t, __bf16,
145 z0 = svmls_n_bf16_z (p0, z0, z1, d4),
146 z0 = svmls_z (p0, z0, z1, d4))
147
148 /*
149 ** mls_h4_bf16_z_tied2:
150 ** mov (z[0-9]+\.h), h4
151 ** mov (z[0-9]+)\.d, z0\.d
152 ** movprfx z0\.h, p0/z, z1\.h
153 ** bfmls z0\.h, p0/m, \2\.h, \1
154 ** ret
155 */
156 TEST_UNIFORM_ZD (mls_h4_bf16_z_tied2, svbfloat16_t, __bf16,
157 z0 = svmls_n_bf16_z (p0, z1, z0, d4),
158 z0 = svmls_z (p0, z1, z0, d4))
159
160 /*
161 ** mls_h4_bf16_z_untied:
162 ** mov (z[0-9]+\.h), h4
163 ** movprfx z0\.h, p0/z, z1\.h
164 ** bfmls z0\.h, p0/m, z2\.h, \1
165 ** ret
166 */
167 TEST_UNIFORM_ZD (mls_h4_bf16_z_untied, svbfloat16_t, __bf16,
168 z0 = svmls_n_bf16_z (p0, z1, z2, d4),
169 z0 = svmls_z (p0, z1, z2, d4))
170
171 /*
172 ** mls_2_bf16_z_tied1:
173 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
174 ** movprfx z0\.h, p0/z, z0\.h
175 ** bfmls z0\.h, p0/m, z1\.h, \1
176 ** ret
177 */
178 TEST_UNIFORM_Z (mls_2_bf16_z_tied1, svbfloat16_t,
179 z0 = svmls_n_bf16_z (p0, z0, z1, 2),
180 z0 = svmls_z (p0, z0, z1, 2))
181
182 /*
183 ** mls_bf16_x_tied1:
184 ** bfmls z0\.h, p0/m, z1\.h, z2\.h
185 ** ret
186 */
187 TEST_UNIFORM_Z (mls_bf16_x_tied1, svbfloat16_t,
188 z0 = svmls_bf16_x (p0, z0, z1, z2),
189 z0 = svmls_x (p0, z0, z1, z2))
190
191 /*
192 ** mls_bf16_x_tied2:
193 ** mov (z[0-9]+)\.d, z0\.d
194 ** movprfx z0, z1
195 ** bfmls z0\.h, p0/m, \1\.h, z2\.h
196 ** ret
197 */
198 TEST_UNIFORM_Z (mls_bf16_x_tied2, svbfloat16_t,
199 z0 = svmls_bf16_x (p0, z1, z0, z2),
200 z0 = svmls_x (p0, z1, z0, z2))
201
202 /*
203 ** mls_bf16_x_tied3:
204 ** mov (z[0-9]+)\.d, z0\.d
205 ** movprfx z0, z1
206 ** bfmls z0\.h, p0/m, z2\.h, \1\.h
207 ** ret
208 */
209 TEST_UNIFORM_Z (mls_bf16_x_tied3, svbfloat16_t,
210 z0 = svmls_bf16_x (p0, z1, z2, z0),
211 z0 = svmls_x (p0, z1, z2, z0))
212
213 /*
214 ** mls_bf16_x_untied:
215 ** movprfx z0, z1
216 ** bfmls z0\.h, p0/m, z2\.h, z3\.h
217 ** ret
218 */
219 TEST_UNIFORM_Z (mls_bf16_x_untied, svbfloat16_t,
220 z0 = svmls_bf16_x (p0, z1, z2, z3),
221 z0 = svmls_x (p0, z1, z2, z3))
222
223 /*
224 ** mls_h4_bf16_x_tied1:
225 ** mov (z[0-9]+\.h), h4
226 ** bfmls z0\.h, p0/m, z1\.h, \1
227 ** ret
228 */
229 TEST_UNIFORM_ZD (mls_h4_bf16_x_tied1, svbfloat16_t, __bf16,
230 z0 = svmls_n_bf16_x (p0, z0, z1, d4),
231 z0 = svmls_x (p0, z0, z1, d4))
232
233 /*
234 ** mls_h4_bf16_x_tied2:
235 ** mov (z[0-9]+\.h), h4
236 ** mov (z[0-9]+)\.d, z0\.d
237 ** movprfx z0, z1
238 ** bfmls z0\.h, p0/m, \2\.h, \1
239 ** ret
240 */
241 TEST_UNIFORM_ZD (mls_h4_bf16_x_tied2, svbfloat16_t, __bf16,
242 z0 = svmls_n_bf16_x (p0, z1, z0, d4),
243 z0 = svmls_x (p0, z1, z0, d4))
244
245 /*
246 ** mls_h4_bf16_x_untied:
247 ** mov (z[0-9]+\.h), h4
248 ** movprfx z0, z1
249 ** bfmls z0\.h, p0/m, z2\.h, \1
250 ** ret
251 */
252 TEST_UNIFORM_ZD (mls_h4_bf16_x_untied, svbfloat16_t, __bf16,
253 z0 = svmls_n_bf16_x (p0, z1, z2, d4),
254 z0 = svmls_x (p0, z1, z2, d4))
255
256 /*
257 ** mls_2_bf16_x_tied1:
258 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
259 ** bfmls z0\.h, p0/m, z1\.h, \1
260 ** ret
261 */
262 TEST_UNIFORM_Z (mls_2_bf16_x_tied1, svbfloat16_t,
263 z0 = svmls_n_bf16_x (p0, z0, z1, 2),
264 z0 = svmls_x (p0, z0, z1, 2))
265
266 /*
267 ** ptrue_mls_bf16_x_tied1:
268 ** ...
269 ** ptrue p[0-9]+\.b[^\n]*
270 ** ...
271 ** ret
272 */
273 TEST_UNIFORM_Z (ptrue_mls_bf16_x_tied1, svbfloat16_t,
274 z0 = svmls_bf16_x (svptrue_b16 (), z0, z1, z2),
275 z0 = svmls_x (svptrue_b16 (), z0, z1, z2))
276
277 /*
278 ** ptrue_mls_bf16_x_tied2:
279 ** ...
280 ** ptrue p[0-9]+\.b[^\n]*
281 ** ...
282 ** ret
283 */
284 TEST_UNIFORM_Z (ptrue_mls_bf16_x_tied2, svbfloat16_t,
285 z0 = svmls_bf16_x (svptrue_b16 (), z1, z0, z2),
286 z0 = svmls_x (svptrue_b16 (), z1, z0, z2))
287
288 /*
289 ** ptrue_mls_bf16_x_tied3:
290 ** ...
291 ** ptrue p[0-9]+\.b[^\n]*
292 ** ...
293 ** ret
294 */
295 TEST_UNIFORM_Z (ptrue_mls_bf16_x_tied3, svbfloat16_t,
296 z0 = svmls_bf16_x (svptrue_b16 (), z1, z2, z0),
297 z0 = svmls_x (svptrue_b16 (), z1, z2, z0))
298
299 /*
300 ** ptrue_mls_bf16_x_untied:
301 ** ...
302 ** ptrue p[0-9]+\.b[^\n]*
303 ** ...
304 ** ret
305 */
306 TEST_UNIFORM_Z (ptrue_mls_bf16_x_untied, svbfloat16_t,
307 z0 = svmls_bf16_x (svptrue_b16 (), z1, z2, z3),
308 z0 = svmls_x (svptrue_b16 (), z1, z2, z3))
309
310 /*
311 ** ptrue_mls_2_bf16_x_tied1:
312 ** ...
313 ** ptrue p[0-9]+\.b[^\n]*
314 ** ...
315 ** ret
316 */
317 TEST_UNIFORM_Z (ptrue_mls_2_bf16_x_tied1, svbfloat16_t,
318 z0 = svmls_n_bf16_x (svptrue_b16 (), z0, z1, 2),
319 z0 = svmls_x (svptrue_b16 (), z0, z1, 2))
320
321 /*
322 ** ptrue_mls_2_bf16_x_tied2:
323 ** ...
324 ** ptrue p[0-9]+\.b[^\n]*
325 ** ...
326 ** ret
327 */
328 TEST_UNIFORM_Z (ptrue_mls_2_bf16_x_tied2, svbfloat16_t,
329 z0 = svmls_n_bf16_x (svptrue_b16 (), z1, z0, 2),
330 z0 = svmls_x (svptrue_b16 (), z1, z0, 2))
331
332 /*
333 ** ptrue_mls_2_bf16_x_untied:
334 ** ...
335 ** ptrue p[0-9]+\.b[^\n]*
336 ** ...
337 ** ret
338 */
339 TEST_UNIFORM_Z (ptrue_mls_2_bf16_x_untied, svbfloat16_t,
340 z0 = svmls_n_bf16_x (svptrue_b16 (), z1, z2, 2),
341 z0 = svmls_x (svptrue_b16 (), z1, z2, 2))