1 /* { dg-additional-options "-fno-trapping-math" } */
2 /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */
3
4 #include "test_sve_acle.h"
5
6 /*
7 ** add_f16_m_tied1:
8 ** fadd z0\.h, p0/m, z0\.h, z1\.h
9 ** ret
10 */
11 TEST_UNIFORM_Z (add_f16_m_tied1, svfloat16_t,
12 z0 = svadd_f16_m (p0, z0, z1),
13 z0 = svadd_m (p0, z0, z1))
14
15 /*
16 ** add_f16_m_tied2:
17 ** mov (z[0-9]+)\.d, z0\.d
18 ** movprfx z0, z1
19 ** fadd z0\.h, p0/m, z0\.h, \1\.h
20 ** ret
21 */
22 TEST_UNIFORM_Z (add_f16_m_tied2, svfloat16_t,
23 z0 = svadd_f16_m (p0, z1, z0),
24 z0 = svadd_m (p0, z1, z0))
25
26 /*
27 ** add_f16_m_untied:
28 ** movprfx z0, z1
29 ** fadd z0\.h, p0/m, z0\.h, z2\.h
30 ** ret
31 */
32 TEST_UNIFORM_Z (add_f16_m_untied, svfloat16_t,
33 z0 = svadd_f16_m (p0, z1, z2),
34 z0 = svadd_m (p0, z1, z2))
35
36 /*
37 ** add_h4_f16_m_tied1:
38 ** mov (z[0-9]+\.h), h4
39 ** fadd z0\.h, p0/m, z0\.h, \1
40 ** ret
41 */
42 TEST_UNIFORM_ZD (add_h4_f16_m_tied1, svfloat16_t, __fp16,
43 z0 = svadd_n_f16_m (p0, z0, d4),
44 z0 = svadd_m (p0, z0, d4))
45
46 /*
47 ** add_h4_f16_m_untied:
48 ** mov (z[0-9]+\.h), h4
49 ** movprfx z0, z1
50 ** fadd z0\.h, p0/m, z0\.h, \1
51 ** ret
52 */
53 TEST_UNIFORM_ZD (add_h4_f16_m_untied, svfloat16_t, __fp16,
54 z0 = svadd_n_f16_m (p0, z1, d4),
55 z0 = svadd_m (p0, z1, d4))
56
57 /*
58 ** add_1_f16_m_tied1:
59 ** fadd z0\.h, p0/m, z0\.h, #1\.0
60 ** ret
61 */
62 TEST_UNIFORM_Z (add_1_f16_m_tied1, svfloat16_t,
63 z0 = svadd_n_f16_m (p0, z0, 1),
64 z0 = svadd_m (p0, z0, 1))
65
66 /*
67 ** add_1_f16_m_untied:
68 ** movprfx z0, z1
69 ** fadd z0\.h, p0/m, z0\.h, #1\.0
70 ** ret
71 */
72 TEST_UNIFORM_Z (add_1_f16_m_untied, svfloat16_t,
73 z0 = svadd_n_f16_m (p0, z1, 1),
74 z0 = svadd_m (p0, z1, 1))
75
76 /*
77 ** add_0p5_f16_m_tied1:
78 ** fadd z0\.h, p0/m, z0\.h, #0\.5
79 ** ret
80 */
81 TEST_UNIFORM_Z (add_0p5_f16_m_tied1, svfloat16_t,
82 z0 = svadd_n_f16_m (p0, z0, 0.5),
83 z0 = svadd_m (p0, z0, 0.5))
84
85 /*
86 ** add_0p5_f16_m_untied:
87 ** movprfx z0, z1
88 ** fadd z0\.h, p0/m, z0\.h, #0\.5
89 ** ret
90 */
91 TEST_UNIFORM_Z (add_0p5_f16_m_untied, svfloat16_t,
92 z0 = svadd_n_f16_m (p0, z1, 0.5),
93 z0 = svadd_m (p0, z1, 0.5))
94
95 /*
96 ** add_m1_f16_m_tied1:
97 ** fsub z0\.h, p0/m, z0\.h, #1\.0
98 ** ret
99 */
100 TEST_UNIFORM_Z (add_m1_f16_m_tied1, svfloat16_t,
101 z0 = svadd_n_f16_m (p0, z0, -1),
102 z0 = svadd_m (p0, z0, -1))
103
104 /*
105 ** add_m1_f16_m_untied:
106 ** movprfx z0, z1
107 ** fsub z0\.h, p0/m, z0\.h, #1\.0
108 ** ret
109 */
110 TEST_UNIFORM_Z (add_m1_f16_m_untied, svfloat16_t,
111 z0 = svadd_n_f16_m (p0, z1, -1),
112 z0 = svadd_m (p0, z1, -1))
113
114 /*
115 ** add_m0p5_f16_m_tied1:
116 ** fsub z0\.h, p0/m, z0\.h, #0\.5
117 ** ret
118 */
119 TEST_UNIFORM_Z (add_m0p5_f16_m_tied1, svfloat16_t,
120 z0 = svadd_n_f16_m (p0, z0, -0.5),
121 z0 = svadd_m (p0, z0, -0.5))
122
123 /*
124 ** add_m0p5_f16_m_untied:
125 ** movprfx z0, z1
126 ** fsub z0\.h, p0/m, z0\.h, #0\.5
127 ** ret
128 */
129 TEST_UNIFORM_Z (add_m0p5_f16_m_untied, svfloat16_t,
130 z0 = svadd_n_f16_m (p0, z1, -0.5),
131 z0 = svadd_m (p0, z1, -0.5))
132
133 /*
134 ** add_m2_f16_m:
135 ** fmov (z[0-9]+\.h), #-2\.0(?:e\+0)?
136 ** fadd z0\.h, p0/m, z0\.h, \1
137 ** ret
138 */
139 TEST_UNIFORM_Z (add_m2_f16_m, svfloat16_t,
140 z0 = svadd_n_f16_m (p0, z0, -2),
141 z0 = svadd_m (p0, z0, -2))
142
143 /*
144 ** add_f16_z_tied1:
145 ** movprfx z0\.h, p0/z, z0\.h
146 ** fadd z0\.h, p0/m, z0\.h, z1\.h
147 ** ret
148 */
149 TEST_UNIFORM_Z (add_f16_z_tied1, svfloat16_t,
150 z0 = svadd_f16_z (p0, z0, z1),
151 z0 = svadd_z (p0, z0, z1))
152
153 /*
154 ** add_f16_z_tied2:
155 ** movprfx z0\.h, p0/z, z0\.h
156 ** fadd z0\.h, p0/m, z0\.h, z1\.h
157 ** ret
158 */
159 TEST_UNIFORM_Z (add_f16_z_tied2, svfloat16_t,
160 z0 = svadd_f16_z (p0, z1, z0),
161 z0 = svadd_z (p0, z1, z0))
162
163 /*
164 ** add_f16_z_untied:
165 ** (
166 ** movprfx z0\.h, p0/z, z1\.h
167 ** fadd z0\.h, p0/m, z0\.h, z2\.h
168 ** |
169 ** movprfx z0\.h, p0/z, z2\.h
170 ** fadd z0\.h, p0/m, z0\.h, z1\.h
171 ** )
172 ** ret
173 */
174 TEST_UNIFORM_Z (add_f16_z_untied, svfloat16_t,
175 z0 = svadd_f16_z (p0, z1, z2),
176 z0 = svadd_z (p0, z1, z2))
177
178 /*
179 ** add_h4_f16_z_tied1:
180 ** mov (z[0-9]+\.h), h4
181 ** movprfx z0\.h, p0/z, z0\.h
182 ** fadd z0\.h, p0/m, z0\.h, \1
183 ** ret
184 */
185 TEST_UNIFORM_ZD (add_h4_f16_z_tied1, svfloat16_t, __fp16,
186 z0 = svadd_n_f16_z (p0, z0, d4),
187 z0 = svadd_z (p0, z0, d4))
188
189 /*
190 ** add_h4_f16_z_untied:
191 ** mov (z[0-9]+\.h), h4
192 ** (
193 ** movprfx z0\.h, p0/z, z1\.h
194 ** fadd z0\.h, p0/m, z0\.h, \1
195 ** |
196 ** movprfx z0\.h, p0/z, \1
197 ** fadd z0\.h, p0/m, z0\.h, z1\.h
198 ** )
199 ** ret
200 */
201 TEST_UNIFORM_ZD (add_h4_f16_z_untied, svfloat16_t, __fp16,
202 z0 = svadd_n_f16_z (p0, z1, d4),
203 z0 = svadd_z (p0, z1, d4))
204
205 /*
206 ** add_1_f16_z_tied1:
207 ** movprfx z0\.h, p0/z, z0\.h
208 ** fadd z0\.h, p0/m, z0\.h, #1\.0
209 ** ret
210 */
211 TEST_UNIFORM_Z (add_1_f16_z_tied1, svfloat16_t,
212 z0 = svadd_n_f16_z (p0, z0, 1),
213 z0 = svadd_z (p0, z0, 1))
214
215 /*
216 ** add_1_f16_z_untied:
217 ** movprfx z0\.h, p0/z, z1\.h
218 ** fadd z0\.h, p0/m, z0\.h, #1\.0
219 ** ret
220 */
221 TEST_UNIFORM_Z (add_1_f16_z_untied, svfloat16_t,
222 z0 = svadd_n_f16_z (p0, z1, 1),
223 z0 = svadd_z (p0, z1, 1))
224
225 /*
226 ** add_0p5_f16_z_tied1:
227 ** movprfx z0\.h, p0/z, z0\.h
228 ** fadd z0\.h, p0/m, z0\.h, #0\.5
229 ** ret
230 */
231 TEST_UNIFORM_Z (add_0p5_f16_z_tied1, svfloat16_t,
232 z0 = svadd_n_f16_z (p0, z0, 0.5),
233 z0 = svadd_z (p0, z0, 0.5))
234
235 /*
236 ** add_0p5_f16_z_untied:
237 ** movprfx z0\.h, p0/z, z1\.h
238 ** fadd z0\.h, p0/m, z0\.h, #0\.5
239 ** ret
240 */
241 TEST_UNIFORM_Z (add_0p5_f16_z_untied, svfloat16_t,
242 z0 = svadd_n_f16_z (p0, z1, 0.5),
243 z0 = svadd_z (p0, z1, 0.5))
244
245 /*
246 ** add_m1_f16_z_tied1:
247 ** movprfx z0\.h, p0/z, z0\.h
248 ** fsub z0\.h, p0/m, z0\.h, #1\.0
249 ** ret
250 */
251 TEST_UNIFORM_Z (add_m1_f16_z_tied1, svfloat16_t,
252 z0 = svadd_n_f16_z (p0, z0, -1),
253 z0 = svadd_z (p0, z0, -1))
254
255 /*
256 ** add_m1_f16_z_untied:
257 ** movprfx z0\.h, p0/z, z1\.h
258 ** fsub z0\.h, p0/m, z0\.h, #1\.0
259 ** ret
260 */
261 TEST_UNIFORM_Z (add_m1_f16_z_untied, svfloat16_t,
262 z0 = svadd_n_f16_z (p0, z1, -1),
263 z0 = svadd_z (p0, z1, -1))
264
265 /*
266 ** add_m0p5_f16_z_tied1:
267 ** movprfx z0\.h, p0/z, z0\.h
268 ** fsub z0\.h, p0/m, z0\.h, #0\.5
269 ** ret
270 */
271 TEST_UNIFORM_Z (add_m0p5_f16_z_tied1, svfloat16_t,
272 z0 = svadd_n_f16_z (p0, z0, -0.5),
273 z0 = svadd_z (p0, z0, -0.5))
274
275 /*
276 ** add_m0p5_f16_z_untied:
277 ** movprfx z0\.h, p0/z, z1\.h
278 ** fsub z0\.h, p0/m, z0\.h, #0\.5
279 ** ret
280 */
281 TEST_UNIFORM_Z (add_m0p5_f16_z_untied, svfloat16_t,
282 z0 = svadd_n_f16_z (p0, z1, -0.5),
283 z0 = svadd_z (p0, z1, -0.5))
284
285 /*
286 ** add_m2_f16_z:
287 ** fmov (z[0-9]+\.h), #-2\.0(?:e\+0)?
288 ** movprfx z0\.h, p0/z, z0\.h
289 ** fadd z0\.h, p0/m, z0\.h, \1
290 ** ret
291 */
292 TEST_UNIFORM_Z (add_m2_f16_z, svfloat16_t,
293 z0 = svadd_n_f16_z (p0, z0, -2),
294 z0 = svadd_z (p0, z0, -2))
295
296 /*
297 ** add_f16_x_tied1:
298 ** fadd z0\.h, (z0\.h, z1\.h|z1\.h, z0\.h)
299 ** ret
300 */
301 TEST_UNIFORM_Z (add_f16_x_tied1, svfloat16_t,
302 z0 = svadd_f16_x (p0, z0, z1),
303 z0 = svadd_x (p0, z0, z1))
304
305 /*
306 ** add_f16_x_tied2:
307 ** fadd z0\.h, (z0\.h, z1\.h|z1\.h, z0\.h)
308 ** ret
309 */
310 TEST_UNIFORM_Z (add_f16_x_tied2, svfloat16_t,
311 z0 = svadd_f16_x (p0, z1, z0),
312 z0 = svadd_x (p0, z1, z0))
313
314 /*
315 ** add_f16_x_untied:
316 ** fadd z0\.h, (z1\.h, z2\.h|z2\.h, z1\.h)
317 ** ret
318 */
319 TEST_UNIFORM_Z (add_f16_x_untied, svfloat16_t,
320 z0 = svadd_f16_x (p0, z1, z2),
321 z0 = svadd_x (p0, z1, z2))
322
323 /*
324 ** add_h4_f16_x_tied1:
325 ** mov (z[0-9]+\.h), h4
326 ** fadd z0\.h, (z0\.h, \1|\1, z0\.h)
327 ** ret
328 */
329 TEST_UNIFORM_ZD (add_h4_f16_x_tied1, svfloat16_t, __fp16,
330 z0 = svadd_n_f16_x (p0, z0, d4),
331 z0 = svadd_x (p0, z0, d4))
332
333 /*
334 ** add_h4_f16_x_untied:
335 ** mov (z[0-9]+\.h), h4
336 ** fadd z0\.h, (z1\.h, \1|\1, z1\.h)
337 ** ret
338 */
339 TEST_UNIFORM_ZD (add_h4_f16_x_untied, svfloat16_t, __fp16,
340 z0 = svadd_n_f16_x (p0, z1, d4),
341 z0 = svadd_x (p0, z1, d4))
342
343 /*
344 ** add_1_f16_x_tied1:
345 ** fadd z0\.h, p0/m, z0\.h, #1\.0
346 ** ret
347 */
348 TEST_UNIFORM_Z (add_1_f16_x_tied1, svfloat16_t,
349 z0 = svadd_n_f16_x (p0, z0, 1),
350 z0 = svadd_x (p0, z0, 1))
351
352 /*
353 ** add_1_f16_x_untied:
354 ** movprfx z0, z1
355 ** fadd z0\.h, p0/m, z0\.h, #1\.0
356 ** ret
357 */
358 TEST_UNIFORM_Z (add_1_f16_x_untied, svfloat16_t,
359 z0 = svadd_n_f16_x (p0, z1, 1),
360 z0 = svadd_x (p0, z1, 1))
361
362 /*
363 ** add_0p5_f16_x_tied1:
364 ** fadd z0\.h, p0/m, z0\.h, #0\.5
365 ** ret
366 */
367 TEST_UNIFORM_Z (add_0p5_f16_x_tied1, svfloat16_t,
368 z0 = svadd_n_f16_x (p0, z0, 0.5),
369 z0 = svadd_x (p0, z0, 0.5))
370
371 /*
372 ** add_0p5_f16_x_untied:
373 ** movprfx z0, z1
374 ** fadd z0\.h, p0/m, z0\.h, #0\.5
375 ** ret
376 */
377 TEST_UNIFORM_Z (add_0p5_f16_x_untied, svfloat16_t,
378 z0 = svadd_n_f16_x (p0, z1, 0.5),
379 z0 = svadd_x (p0, z1, 0.5))
380
381 /*
382 ** add_m1_f16_x_tied1:
383 ** fsub z0\.h, p0/m, z0\.h, #1\.0
384 ** ret
385 */
386 TEST_UNIFORM_Z (add_m1_f16_x_tied1, svfloat16_t,
387 z0 = svadd_n_f16_x (p0, z0, -1),
388 z0 = svadd_x (p0, z0, -1))
389
390 /*
391 ** add_m1_f16_x_untied:
392 ** movprfx z0, z1
393 ** fsub z0\.h, p0/m, z0\.h, #1\.0
394 ** ret
395 */
396 TEST_UNIFORM_Z (add_m1_f16_x_untied, svfloat16_t,
397 z0 = svadd_n_f16_x (p0, z1, -1),
398 z0 = svadd_x (p0, z1, -1))
399
400 /*
401 ** add_m0p5_f16_x_tied1:
402 ** fsub z0\.h, p0/m, z0\.h, #0\.5
403 ** ret
404 */
405 TEST_UNIFORM_Z (add_m0p5_f16_x_tied1, svfloat16_t,
406 z0 = svadd_n_f16_x (p0, z0, -0.5),
407 z0 = svadd_x (p0, z0, -0.5))
408
409 /*
410 ** add_m0p5_f16_x_untied:
411 ** movprfx z0, z1
412 ** fsub z0\.h, p0/m, z0\.h, #0\.5
413 ** ret
414 */
415 TEST_UNIFORM_Z (add_m0p5_f16_x_untied, svfloat16_t,
416 z0 = svadd_n_f16_x (p0, z1, -0.5),
417 z0 = svadd_x (p0, z1, -0.5))
418
419 /*
420 ** add_2_f16_x_tied1:
421 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
422 ** fadd z0\.h, (z0\.h, \1|\1, z0\.h)
423 ** ret
424 */
425 TEST_UNIFORM_Z (add_2_f16_x_tied1, svfloat16_t,
426 z0 = svadd_n_f16_x (p0, z0, 2),
427 z0 = svadd_x (p0, z0, 2))
428
429 /*
430 ** add_2_f16_x_untied:
431 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
432 ** fadd z0\.h, (z1\.h, \1|\1, z1\.h)
433 ** ret
434 */
435 TEST_UNIFORM_Z (add_2_f16_x_untied, svfloat16_t,
436 z0 = svadd_n_f16_x (p0, z1, 2),
437 z0 = svadd_x (p0, z1, 2))
438
439 /*
440 ** ptrue_add_f16_x_tied1:
441 ** fadd z0\.h, (z0\.h, z1\.h|z1\.h, z0\.h)
442 ** ret
443 */
444 TEST_UNIFORM_Z (ptrue_add_f16_x_tied1, svfloat16_t,
445 z0 = svadd_f16_x (svptrue_b16 (), z0, z1),
446 z0 = svadd_x (svptrue_b16 (), z0, z1))
447
448 /*
449 ** ptrue_add_f16_x_tied2:
450 ** fadd z0\.h, (z0\.h, z1\.h|z1\.h, z0\.h)
451 ** ret
452 */
453 TEST_UNIFORM_Z (ptrue_add_f16_x_tied2, svfloat16_t,
454 z0 = svadd_f16_x (svptrue_b16 (), z1, z0),
455 z0 = svadd_x (svptrue_b16 (), z1, z0))
456
457 /*
458 ** ptrue_add_f16_x_untied:
459 ** fadd z0\.h, (z1\.h, z2\.h|z2\.h, z1\.h)
460 ** ret
461 */
462 TEST_UNIFORM_Z (ptrue_add_f16_x_untied, svfloat16_t,
463 z0 = svadd_f16_x (svptrue_b16 (), z1, z2),
464 z0 = svadd_x (svptrue_b16 (), z1, z2))
465
466 /*
467 ** ptrue_add_1_f16_x_tied1:
468 ** ...
469 ** ptrue p[0-9]+\.b[^\n]*
470 ** ...
471 ** ret
472 */
473 TEST_UNIFORM_Z (ptrue_add_1_f16_x_tied1, svfloat16_t,
474 z0 = svadd_n_f16_x (svptrue_b16 (), z0, 1),
475 z0 = svadd_x (svptrue_b16 (), z0, 1))
476
477 /*
478 ** ptrue_add_1_f16_x_untied:
479 ** ...
480 ** ptrue p[0-9]+\.b[^\n]*
481 ** ...
482 ** ret
483 */
484 TEST_UNIFORM_Z (ptrue_add_1_f16_x_untied, svfloat16_t,
485 z0 = svadd_n_f16_x (svptrue_b16 (), z1, 1),
486 z0 = svadd_x (svptrue_b16 (), z1, 1))
487
488 /*
489 ** ptrue_add_0p5_f16_x_tied1:
490 ** ...
491 ** ptrue p[0-9]+\.b[^\n]*
492 ** ...
493 ** ret
494 */
495 TEST_UNIFORM_Z (ptrue_add_0p5_f16_x_tied1, svfloat16_t,
496 z0 = svadd_n_f16_x (svptrue_b16 (), z0, 0.5),
497 z0 = svadd_x (svptrue_b16 (), z0, 0.5))
498
499 /*
500 ** ptrue_add_0p5_f16_x_untied:
501 ** ...
502 ** ptrue p[0-9]+\.b[^\n]*
503 ** ...
504 ** ret
505 */
506 TEST_UNIFORM_Z (ptrue_add_0p5_f16_x_untied, svfloat16_t,
507 z0 = svadd_n_f16_x (svptrue_b16 (), z1, 0.5),
508 z0 = svadd_x (svptrue_b16 (), z1, 0.5))
509
510 /*
511 ** ptrue_add_m1_f16_x_tied1:
512 ** ...
513 ** ptrue p[0-9]+\.b[^\n]*
514 ** ...
515 ** ret
516 */
517 TEST_UNIFORM_Z (ptrue_add_m1_f16_x_tied1, svfloat16_t,
518 z0 = svadd_n_f16_x (svptrue_b16 (), z0, -1),
519 z0 = svadd_x (svptrue_b16 (), z0, -1))
520
521 /*
522 ** ptrue_add_m1_f16_x_untied:
523 ** ...
524 ** ptrue p[0-9]+\.b[^\n]*
525 ** ...
526 ** ret
527 */
528 TEST_UNIFORM_Z (ptrue_add_m1_f16_x_untied, svfloat16_t,
529 z0 = svadd_n_f16_x (svptrue_b16 (), z1, -1),
530 z0 = svadd_x (svptrue_b16 (), z1, -1))
531
532 /*
533 ** ptrue_add_m0p5_f16_x_tied1:
534 ** ...
535 ** ptrue p[0-9]+\.b[^\n]*
536 ** ...
537 ** ret
538 */
539 TEST_UNIFORM_Z (ptrue_add_m0p5_f16_x_tied1, svfloat16_t,
540 z0 = svadd_n_f16_x (svptrue_b16 (), z0, -0.5),
541 z0 = svadd_x (svptrue_b16 (), z0, -0.5))
542
543 /*
544 ** ptrue_add_m0p5_f16_x_untied:
545 ** ...
546 ** ptrue p[0-9]+\.b[^\n]*
547 ** ...
548 ** ret
549 */
550 TEST_UNIFORM_Z (ptrue_add_m0p5_f16_x_untied, svfloat16_t,
551 z0 = svadd_n_f16_x (svptrue_b16 (), z1, -0.5),
552 z0 = svadd_x (svptrue_b16 (), z1, -0.5))
553
554 /*
555 ** ptrue_add_2_f16_x_tied1:
556 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
557 ** fadd z0\.h, (z0\.h, \1|\1, z0\.h)
558 ** ret
559 */
560 TEST_UNIFORM_Z (ptrue_add_2_f16_x_tied1, svfloat16_t,
561 z0 = svadd_n_f16_x (svptrue_b16 (), z0, 2),
562 z0 = svadd_x (svptrue_b16 (), z0, 2))
563
564 /*
565 ** ptrue_add_2_f16_x_untied:
566 ** fmov (z[0-9]+\.h), #2\.0(?:e\+0)?
567 ** fadd z0\.h, (z1\.h, \1|\1, z1\.h)
568 ** ret
569 */
570 TEST_UNIFORM_Z (ptrue_add_2_f16_x_untied, svfloat16_t,
571 z0 = svadd_n_f16_x (svptrue_b16 (), z1, 2),
572 z0 = svadd_x (svptrue_b16 (), z1, 2))