1 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
2
3 #include "test_sve_acle.h"
4
5 /*
6 ** nmad_f16_m_tied1:
7 ** fnmad z0\.h, p0/m, z1\.h, z2\.h
8 ** ret
9 */
10 TEST_UNIFORM_Z (nmad_f16_m_tied1, svfloat16_t,
11 z0 = svnmad_f16_m (p0, z0, z1, z2),
12 z0 = svnmad_m (p0, z0, z1, z2))
13
14 /*
15 ** nmad_f16_m_tied2:
16 ** mov (z[0-9]+)\.d, z0\.d
17 ** movprfx z0, z1
18 ** fnmad z0\.h, p0/m, \1\.h, z2\.h
19 ** ret
20 */
21 TEST_UNIFORM_Z (nmad_f16_m_tied2, svfloat16_t,
22 z0 = svnmad_f16_m (p0, z1, z0, z2),
23 z0 = svnmad_m (p0, z1, z0, z2))
24
25 /*
26 ** nmad_f16_m_tied3:
27 ** mov (z[0-9]+)\.d, z0\.d
28 ** movprfx z0, z1
29 ** fnmad z0\.h, p0/m, z2\.h, \1\.h
30 ** ret
31 */
32 TEST_UNIFORM_Z (nmad_f16_m_tied3, svfloat16_t,
33 z0 = svnmad_f16_m (p0, z1, z2, z0),
34 z0 = svnmad_m (p0, z1, z2, z0))
35
36 /*
37 ** nmad_f16_m_untied:
38 ** movprfx z0, z1
39 ** fnmad z0\.h, p0/m, z2\.h, z3\.h
40 ** ret
41 */
42 TEST_UNIFORM_Z (nmad_f16_m_untied, svfloat16_t,
43 z0 = svnmad_f16_m (p0, z1, z2, z3),
44 z0 = svnmad_m (p0, z1, z2, z3))
45
46 /*
47 ** nmad_h4_f16_m_tied1:
48 ** mov (z[0-9]+\.h), h4
49 ** fnmad z0\.h, p0/m, z1\.h, \1
50 ** ret
51 */
52 TEST_UNIFORM_ZD (nmad_h4_f16_m_tied1, svfloat16_t, __fp16,
53 z0 = svnmad_n_f16_m (p0, z0, z1, d4),
54 z0 = svnmad_m (p0, z0, z1, d4))
55
56 /*
57 ** nmad_h4_f16_m_untied:
58 ** mov (z[0-9]+\.h), h4
59 ** movprfx z0, z1
60 ** fnmad z0\.h, p0/m, z2\.h, \1
61 ** ret
62 */
63 TEST_UNIFORM_ZD (nmad_h4_f16_m_untied, svfloat16_t, __fp16,
64 z0 = svnmad_n_f16_m (p0, z1, z2, d4),
65 z0 = svnmad_m (p0, z1, z2, d4))
66
67 /*
68 ** nmad_2_f16_m_tied1:
69 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
70 ** fnmad z0\.h, p0/m, z1\.h, \1
71 ** ret
72 */
73 TEST_UNIFORM_Z (nmad_2_f16_m_tied1, svfloat16_t,
74 z0 = svnmad_n_f16_m (p0, z0, z1, 2),
75 z0 = svnmad_m (p0, z0, z1, 2))
76
77 /*
78 ** nmad_2_f16_m_untied: { xfail *-*-* }
79 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
80 ** movprfx z0, z1
81 ** fnmad z0\.h, p0/m, z2\.h, \1
82 ** ret
83 */
84 TEST_UNIFORM_Z (nmad_2_f16_m_untied, svfloat16_t,
85 z0 = svnmad_n_f16_m (p0, z1, z2, 2),
86 z0 = svnmad_m (p0, z1, z2, 2))
87
88 /*
89 ** nmad_f16_z_tied1:
90 ** movprfx z0\.h, p0/z, z0\.h
91 ** fnmad z0\.h, p0/m, z1\.h, z2\.h
92 ** ret
93 */
94 TEST_UNIFORM_Z (nmad_f16_z_tied1, svfloat16_t,
95 z0 = svnmad_f16_z (p0, z0, z1, z2),
96 z0 = svnmad_z (p0, z0, z1, z2))
97
98 /*
99 ** nmad_f16_z_tied2:
100 ** movprfx z0\.h, p0/z, z0\.h
101 ** fnmad z0\.h, p0/m, z1\.h, z2\.h
102 ** ret
103 */
104 TEST_UNIFORM_Z (nmad_f16_z_tied2, svfloat16_t,
105 z0 = svnmad_f16_z (p0, z1, z0, z2),
106 z0 = svnmad_z (p0, z1, z0, z2))
107
108 /*
109 ** nmad_f16_z_tied3:
110 ** movprfx z0\.h, p0/z, z0\.h
111 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
112 ** ret
113 */
114 TEST_UNIFORM_Z (nmad_f16_z_tied3, svfloat16_t,
115 z0 = svnmad_f16_z (p0, z1, z2, z0),
116 z0 = svnmad_z (p0, z1, z2, z0))
117
118 /*
119 ** nmad_f16_z_untied:
120 ** (
121 ** movprfx z0\.h, p0/z, z1\.h
122 ** fnmad z0\.h, p0/m, z2\.h, z3\.h
123 ** |
124 ** movprfx z0\.h, p0/z, z2\.h
125 ** fnmad z0\.h, p0/m, z1\.h, z3\.h
126 ** |
127 ** movprfx z0\.h, p0/z, z3\.h
128 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
129 ** )
130 ** ret
131 */
132 TEST_UNIFORM_Z (nmad_f16_z_untied, svfloat16_t,
133 z0 = svnmad_f16_z (p0, z1, z2, z3),
134 z0 = svnmad_z (p0, z1, z2, z3))
135
136 /*
137 ** nmad_h4_f16_z_tied1:
138 ** mov (z[0-9]+\.h), h4
139 ** movprfx z0\.h, p0/z, z0\.h
140 ** fnmad z0\.h, p0/m, z1\.h, \1
141 ** ret
142 */
143 TEST_UNIFORM_ZD (nmad_h4_f16_z_tied1, svfloat16_t, __fp16,
144 z0 = svnmad_n_f16_z (p0, z0, z1, d4),
145 z0 = svnmad_z (p0, z0, z1, d4))
146
147 /*
148 ** nmad_h4_f16_z_tied2:
149 ** mov (z[0-9]+\.h), h4
150 ** movprfx z0\.h, p0/z, z0\.h
151 ** fnmad z0\.h, p0/m, z1\.h, \1
152 ** ret
153 */
154 TEST_UNIFORM_ZD (nmad_h4_f16_z_tied2, svfloat16_t, __fp16,
155 z0 = svnmad_n_f16_z (p0, z1, z0, d4),
156 z0 = svnmad_z (p0, z1, z0, d4))
157
158 /*
159 ** nmad_h4_f16_z_untied:
160 ** mov (z[0-9]+\.h), h4
161 ** (
162 ** movprfx z0\.h, p0/z, z1\.h
163 ** fnmad z0\.h, p0/m, z2\.h, \1
164 ** |
165 ** movprfx z0\.h, p0/z, z2\.h
166 ** fnmad z0\.h, p0/m, z1\.h, \1
167 ** |
168 ** movprfx z0\.h, p0/z, \1
169 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
170 ** )
171 ** ret
172 */
173 TEST_UNIFORM_ZD (nmad_h4_f16_z_untied, svfloat16_t, __fp16,
174 z0 = svnmad_n_f16_z (p0, z1, z2, d4),
175 z0 = svnmad_z (p0, z1, z2, d4))
176
177 /*
178 ** nmad_2_f16_z_tied1:
179 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
180 ** movprfx z0\.h, p0/z, z0\.h
181 ** fnmad z0\.h, p0/m, z1\.h, \1
182 ** ret
183 */
184 TEST_UNIFORM_Z (nmad_2_f16_z_tied1, svfloat16_t,
185 z0 = svnmad_n_f16_z (p0, z0, z1, 2),
186 z0 = svnmad_z (p0, z0, z1, 2))
187
188 /*
189 ** nmad_2_f16_z_tied2:
190 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
191 ** movprfx z0\.h, p0/z, z0\.h
192 ** fnmad z0\.h, p0/m, z1\.h, \1
193 ** ret
194 */
195 TEST_UNIFORM_Z (nmad_2_f16_z_tied2, svfloat16_t,
196 z0 = svnmad_n_f16_z (p0, z1, z0, 2),
197 z0 = svnmad_z (p0, z1, z0, 2))
198
199 /*
200 ** nmad_2_f16_z_untied:
201 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
202 ** (
203 ** movprfx z0\.h, p0/z, z1\.h
204 ** fnmad z0\.h, p0/m, z2\.h, \1
205 ** |
206 ** movprfx z0\.h, p0/z, z2\.h
207 ** fnmad z0\.h, p0/m, z1\.h, \1
208 ** |
209 ** movprfx z0\.h, p0/z, \1
210 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
211 ** )
212 ** ret
213 */
214 TEST_UNIFORM_Z (nmad_2_f16_z_untied, svfloat16_t,
215 z0 = svnmad_n_f16_z (p0, z1, z2, 2),
216 z0 = svnmad_z (p0, z1, z2, 2))
217
218 /*
219 ** nmad_f16_x_tied1:
220 ** fnmad z0\.h, p0/m, z1\.h, z2\.h
221 ** ret
222 */
223 TEST_UNIFORM_Z (nmad_f16_x_tied1, svfloat16_t,
224 z0 = svnmad_f16_x (p0, z0, z1, z2),
225 z0 = svnmad_x (p0, z0, z1, z2))
226
227 /*
228 ** nmad_f16_x_tied2:
229 ** fnmad z0\.h, p0/m, z1\.h, z2\.h
230 ** ret
231 */
232 TEST_UNIFORM_Z (nmad_f16_x_tied2, svfloat16_t,
233 z0 = svnmad_f16_x (p0, z1, z0, z2),
234 z0 = svnmad_x (p0, z1, z0, z2))
235
236 /*
237 ** nmad_f16_x_tied3:
238 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
239 ** ret
240 */
241 TEST_UNIFORM_Z (nmad_f16_x_tied3, svfloat16_t,
242 z0 = svnmad_f16_x (p0, z1, z2, z0),
243 z0 = svnmad_x (p0, z1, z2, z0))
244
245 /*
246 ** nmad_f16_x_untied:
247 ** (
248 ** movprfx z0, z1
249 ** fnmad z0\.h, p0/m, z2\.h, z3\.h
250 ** |
251 ** movprfx z0, z2
252 ** fnmad z0\.h, p0/m, z1\.h, z3\.h
253 ** |
254 ** movprfx z0, z3
255 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
256 ** )
257 ** ret
258 */
259 TEST_UNIFORM_Z (nmad_f16_x_untied, svfloat16_t,
260 z0 = svnmad_f16_x (p0, z1, z2, z3),
261 z0 = svnmad_x (p0, z1, z2, z3))
262
263 /*
264 ** nmad_h4_f16_x_tied1:
265 ** mov (z[0-9]+\.h), h4
266 ** fnmad z0\.h, p0/m, z1\.h, \1
267 ** ret
268 */
269 TEST_UNIFORM_ZD (nmad_h4_f16_x_tied1, svfloat16_t, __fp16,
270 z0 = svnmad_n_f16_x (p0, z0, z1, d4),
271 z0 = svnmad_x (p0, z0, z1, d4))
272
273 /*
274 ** nmad_h4_f16_x_tied2:
275 ** mov (z[0-9]+\.h), h4
276 ** fnmad z0\.h, p0/m, z1\.h, \1
277 ** ret
278 */
279 TEST_UNIFORM_ZD (nmad_h4_f16_x_tied2, svfloat16_t, __fp16,
280 z0 = svnmad_n_f16_x (p0, z1, z0, d4),
281 z0 = svnmad_x (p0, z1, z0, d4))
282
283 /*
284 ** nmad_h4_f16_x_untied:
285 ** mov z0\.h, h4
286 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
287 ** ret
288 */
289 TEST_UNIFORM_ZD (nmad_h4_f16_x_untied, svfloat16_t, __fp16,
290 z0 = svnmad_n_f16_x (p0, z1, z2, d4),
291 z0 = svnmad_x (p0, z1, z2, d4))
292
293 /*
294 ** nmad_2_f16_x_tied1:
295 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
296 ** fnmad z0\.h, p0/m, z1\.h, \1
297 ** ret
298 */
299 TEST_UNIFORM_Z (nmad_2_f16_x_tied1, svfloat16_t,
300 z0 = svnmad_n_f16_x (p0, z0, z1, 2),
301 z0 = svnmad_x (p0, z0, z1, 2))
302
303 /*
304 ** nmad_2_f16_x_tied2:
305 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
306 ** fnmad z0\.h, p0/m, z1\.h, \1
307 ** ret
308 */
309 TEST_UNIFORM_Z (nmad_2_f16_x_tied2, svfloat16_t,
310 z0 = svnmad_n_f16_x (p0, z1, z0, 2),
311 z0 = svnmad_x (p0, z1, z0, 2))
312
313 /*
314 ** nmad_2_f16_x_untied:
315 ** fmov z0\.h, #2\.0(?:e\+0)?
316 ** fnmla z0\.h, p0/m, z1\.h, z2\.h
317 ** ret
318 */
319 TEST_UNIFORM_Z (nmad_2_f16_x_untied, svfloat16_t,
320 z0 = svnmad_n_f16_x (p0, z1, z2, 2),
321 z0 = svnmad_x (p0, z1, z2, 2))
322
323 /*
324 ** ptrue_nmad_f16_x_tied1:
325 ** ...
326 ** ptrue p[0-9]+\.b[^\n]*
327 ** ...
328 ** ret
329 */
330 TEST_UNIFORM_Z (ptrue_nmad_f16_x_tied1, svfloat16_t,
331 z0 = svnmad_f16_x (svptrue_b16 (), z0, z1, z2),
332 z0 = svnmad_x (svptrue_b16 (), z0, z1, z2))
333
334 /*
335 ** ptrue_nmad_f16_x_tied2:
336 ** ...
337 ** ptrue p[0-9]+\.b[^\n]*
338 ** ...
339 ** ret
340 */
341 TEST_UNIFORM_Z (ptrue_nmad_f16_x_tied2, svfloat16_t,
342 z0 = svnmad_f16_x (svptrue_b16 (), z1, z0, z2),
343 z0 = svnmad_x (svptrue_b16 (), z1, z0, z2))
344
345 /*
346 ** ptrue_nmad_f16_x_tied3:
347 ** ...
348 ** ptrue p[0-9]+\.b[^\n]*
349 ** ...
350 ** ret
351 */
352 TEST_UNIFORM_Z (ptrue_nmad_f16_x_tied3, svfloat16_t,
353 z0 = svnmad_f16_x (svptrue_b16 (), z1, z2, z0),
354 z0 = svnmad_x (svptrue_b16 (), z1, z2, z0))
355
356 /*
357 ** ptrue_nmad_f16_x_untied:
358 ** ...
359 ** ptrue p[0-9]+\.b[^\n]*
360 ** ...
361 ** ret
362 */
363 TEST_UNIFORM_Z (ptrue_nmad_f16_x_untied, svfloat16_t,
364 z0 = svnmad_f16_x (svptrue_b16 (), z1, z2, z3),
365 z0 = svnmad_x (svptrue_b16 (), z1, z2, z3))
366
367 /*
368 ** ptrue_nmad_2_f16_x_tied1:
369 ** ...
370 ** ptrue p[0-9]+\.b[^\n]*
371 ** ...
372 ** ret
373 */
374 TEST_UNIFORM_Z (ptrue_nmad_2_f16_x_tied1, svfloat16_t,
375 z0 = svnmad_n_f16_x (svptrue_b16 (), z0, z1, 2),
376 z0 = svnmad_x (svptrue_b16 (), z0, z1, 2))
377
378 /*
379 ** ptrue_nmad_2_f16_x_tied2:
380 ** ...
381 ** ptrue p[0-9]+\.b[^\n]*
382 ** ...
383 ** ret
384 */
385 TEST_UNIFORM_Z (ptrue_nmad_2_f16_x_tied2, svfloat16_t,
386 z0 = svnmad_n_f16_x (svptrue_b16 (), z1, z0, 2),
387 z0 = svnmad_x (svptrue_b16 (), z1, z0, 2))
388
389 /*
390 ** ptrue_nmad_2_f16_x_untied:
391 ** ...
392 ** ptrue p[0-9]+\.b[^\n]*
393 ** ...
394 ** ret
395 */
396 TEST_UNIFORM_Z (ptrue_nmad_2_f16_x_untied, svfloat16_t,
397 z0 = svnmad_n_f16_x (svptrue_b16 (), z1, z2, 2),
398 z0 = svnmad_x (svptrue_b16 (), z1, z2, 2))