-
Notifications
You must be signed in to change notification settings - Fork 1.5k
/
Copy pathcgemv_t_microk_haswell-4.c
539 lines (424 loc) · 23.9 KB
/
cgemv_t_microk_haswell-4.c
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
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
/***************************************************************************
Copyright (c) 2014, The OpenBLAS Project
All rights reserved.
Redistribution and use in source and binary froms, with or without
modification, are permitted provided that the following conditions are
met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary from must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the
distribution.
3. Neither the name of the OpenBLAS project nor the names of
its contributors may be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*****************************************************************************/
#define HAVE_KERNEL_4x4 1
static void cgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha) __attribute__ ((noinline));
static void cgemv_kernel_4x4( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha)
{
BLASLONG register i = 0;
__asm__ __volatile__
(
"vzeroupper \n\t"
"vxorps %%ymm8 , %%ymm8 , %%ymm8 \n\t" // temp
"vxorps %%ymm9 , %%ymm9 , %%ymm9 \n\t" // temp
"vxorps %%ymm10, %%ymm10, %%ymm10 \n\t" // temp
"vxorps %%ymm11, %%ymm11, %%ymm11 \n\t" // temp
"vxorps %%ymm12, %%ymm12, %%ymm12 \n\t" // temp
"vxorps %%ymm13, %%ymm13, %%ymm13 \n\t"
"vxorps %%ymm14, %%ymm14, %%ymm14 \n\t"
"vxorps %%ymm15, %%ymm15, %%ymm15 \n\t"
"testq $0x04, %1 \n\t"
"jz 2f \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups (%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vmovups (%6,%0,4), %%ymm6 \n\t" // 4 complex values from a2
"vmovups (%7,%0,4), %%ymm7 \n\t" // 4 complex values from a3
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm6 , %%ymm0, %%ymm12 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm6 , %%ymm1, %%ymm13 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm7 , %%ymm0, %%ymm14 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm7 , %%ymm1, %%ymm15 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $8 , %0 \n\t"
"subq $4 , %1 \n\t"
"2: \n\t"
"cmpq $0, %1 \n\t"
"je 3f \n\t"
// ".align 16 \n\t"
"1: \n\t"
"prefetcht0 192(%4,%0,4) \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"prefetcht0 192(%5,%0,4) \n\t"
"vmovups (%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"prefetcht0 192(%2,%0,4) \n\t"
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"prefetcht0 192(%6,%0,4) \n\t"
"vmovups (%6,%0,4), %%ymm6 \n\t" // 4 complex values from a2
"prefetcht0 192(%7,%0,4) \n\t"
"vmovups (%7,%0,4), %%ymm7 \n\t" // 4 complex values from a3
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm6 , %%ymm0, %%ymm12 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm6 , %%ymm1, %%ymm13 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm7 , %%ymm0, %%ymm14 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm7 , %%ymm1, %%ymm15 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vmovups 32(%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups 32(%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"vmovups 32(%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vmovups 32(%6,%0,4), %%ymm6 \n\t" // 4 complex values from a2
"vmovups 32(%7,%0,4), %%ymm7 \n\t" // 4 complex values from a3
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm6 , %%ymm0, %%ymm12 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm6 , %%ymm1, %%ymm13 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm7 , %%ymm0, %%ymm14 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm7 , %%ymm1, %%ymm15 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $16 , %0 \n\t"
"subq $8 , %1 \n\t"
"jnz 1b \n\t"
"3: \n\t"
"vbroadcastss (%8) , %%xmm0 \n\t" // value from alpha
"vbroadcastss 4(%8) , %%xmm1 \n\t" // value from alpha
#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )
"vpermilps $0xb1 , %%ymm9 , %%ymm9 \n\t"
"vpermilps $0xb1 , %%ymm11, %%ymm11 \n\t"
"vpermilps $0xb1 , %%ymm13, %%ymm13 \n\t"
"vpermilps $0xb1 , %%ymm15, %%ymm15 \n\t"
"vaddsubps %%ymm9 , %%ymm8, %%ymm8 \n\t"
"vaddsubps %%ymm11, %%ymm10, %%ymm10 \n\t"
"vaddsubps %%ymm13, %%ymm12, %%ymm12 \n\t"
"vaddsubps %%ymm15, %%ymm14, %%ymm14 \n\t"
#else
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
"vpermilps $0xb1 , %%ymm10, %%ymm10 \n\t"
"vpermilps $0xb1 , %%ymm12, %%ymm12 \n\t"
"vpermilps $0xb1 , %%ymm14, %%ymm14 \n\t"
"vaddsubps %%ymm8 , %%ymm9 , %%ymm8 \n\t"
"vaddsubps %%ymm10, %%ymm11, %%ymm10 \n\t"
"vaddsubps %%ymm12, %%ymm13, %%ymm12 \n\t"
"vaddsubps %%ymm14, %%ymm15, %%ymm14 \n\t"
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
"vpermilps $0xb1 , %%ymm10, %%ymm10 \n\t"
"vpermilps $0xb1 , %%ymm12, %%ymm12 \n\t"
"vpermilps $0xb1 , %%ymm14, %%ymm14 \n\t"
#endif
"vmovsd (%3), %%xmm4 \n\t" // read y
"vmovsd 8(%3), %%xmm5 \n\t"
"vmovsd 16(%3), %%xmm6 \n\t"
"vmovsd 24(%3), %%xmm7 \n\t"
"vextractf128 $1, %%ymm8 , %%xmm9 \n\t"
"vextractf128 $1, %%ymm10, %%xmm11 \n\t"
"vextractf128 $1, %%ymm12, %%xmm13 \n\t"
"vextractf128 $1, %%ymm14, %%xmm15 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vaddps %%xmm12, %%xmm13, %%xmm12 \n\t"
"vaddps %%xmm14, %%xmm15, %%xmm14 \n\t"
"vshufpd $0x1, %%xmm8 , %%xmm8 , %%xmm9 \n\t"
"vshufpd $0x1, %%xmm10, %%xmm10, %%xmm11 \n\t"
"vshufpd $0x1, %%xmm12, %%xmm12, %%xmm13 \n\t"
"vshufpd $0x1, %%xmm14, %%xmm14, %%xmm15 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vaddps %%xmm12, %%xmm13, %%xmm12 \n\t"
"vaddps %%xmm14, %%xmm15, %%xmm14 \n\t"
"vmulps %%xmm8 , %%xmm1 , %%xmm9 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm8 , %%xmm0 , %%xmm8 \n\t" // t_r * alpha_r , t_i * alpha_r
"vmulps %%xmm10, %%xmm1 , %%xmm11 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm10, %%xmm0 , %%xmm10 \n\t" // t_r * alpha_r , t_i * alpha_r
"vmulps %%xmm12, %%xmm1 , %%xmm13 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm12, %%xmm0 , %%xmm12 \n\t" // t_r * alpha_r , t_i * alpha_r
"vmulps %%xmm14, %%xmm1 , %%xmm15 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm14, %%xmm0 , %%xmm14 \n\t" // t_r * alpha_r , t_i * alpha_r
#if !defined(XCONJ)
"vpermilps $0xb1 , %%xmm9 , %%xmm9 \n\t"
"vpermilps $0xb1 , %%xmm11, %%xmm11 \n\t"
"vpermilps $0xb1 , %%xmm13, %%xmm13 \n\t"
"vpermilps $0xb1 , %%xmm15, %%xmm15 \n\t"
"vaddsubps %%xmm9 , %%xmm8, %%xmm8 \n\t"
"vaddsubps %%xmm11, %%xmm10, %%xmm10 \n\t"
"vaddsubps %%xmm13, %%xmm12, %%xmm12 \n\t"
"vaddsubps %%xmm15, %%xmm14, %%xmm14 \n\t"
#else
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
"vpermilps $0xb1 , %%xmm10, %%xmm10 \n\t"
"vpermilps $0xb1 , %%xmm12, %%xmm12 \n\t"
"vpermilps $0xb1 , %%xmm14, %%xmm14 \n\t"
"vaddsubps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddsubps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vaddsubps %%xmm12, %%xmm13, %%xmm12 \n\t"
"vaddsubps %%xmm14, %%xmm15, %%xmm14 \n\t"
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
"vpermilps $0xb1 , %%xmm10, %%xmm10 \n\t"
"vpermilps $0xb1 , %%xmm12, %%xmm12 \n\t"
"vpermilps $0xb1 , %%xmm14, %%xmm14 \n\t"
#endif
"vaddps %%xmm8 , %%xmm4 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm5 , %%xmm10 \n\t"
"vaddps %%xmm12, %%xmm6 , %%xmm12 \n\t"
"vaddps %%xmm14, %%xmm7 , %%xmm14 \n\t"
"vmovsd %%xmm8 , (%3) \n\t"
"vmovsd %%xmm10, 8(%3) \n\t"
"vmovsd %%xmm12, 16(%3) \n\t"
"vmovsd %%xmm14, 24(%3) \n\t"
"vzeroupper \n\t"
:
"+r" (i), // 0
"+r" (n) // 1
:
"r" (x), // 2
"r" (y), // 3
"r" (ap[0]), // 4
"r" (ap[1]), // 5
"r" (ap[2]), // 6
"r" (ap[3]), // 7
"r" (alpha) // 8
: "cc",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);
}
#define HAVE_KERNEL_4x2 1
static void cgemv_kernel_4x2( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha) __attribute__ ((noinline));
static void cgemv_kernel_4x2( BLASLONG n, FLOAT **ap, FLOAT *x, FLOAT *y, FLOAT *alpha)
{
BLASLONG register i = 0;
__asm__ __volatile__
(
"vzeroupper \n\t"
"vxorps %%ymm8 , %%ymm8 , %%ymm8 \n\t" // temp
"vxorps %%ymm9 , %%ymm9 , %%ymm9 \n\t" // temp
"vxorps %%ymm10, %%ymm10, %%ymm10 \n\t" // temp
"vxorps %%ymm11, %%ymm11, %%ymm11 \n\t" // temp
"testq $0x04, %1 \n\t"
"jz 2f \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups (%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $8 , %0 \n\t"
"subq $4 , %1 \n\t"
"2: \n\t"
"cmpq $0, %1 \n\t"
"je 3f \n\t"
// ".align 16 \n\t"
"1: \n\t"
"prefetcht0 192(%4,%0,4) \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"prefetcht0 192(%5,%0,4) \n\t"
"vmovups (%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"prefetcht0 192(%2,%0,4) \n\t"
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vmovups 32(%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups 32(%5,%0,4), %%ymm5 \n\t" // 4 complex values from a1
"vmovups 32(%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vfmadd231ps %%ymm5 , %%ymm0, %%ymm10 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm5 , %%ymm1, %%ymm11 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $16 , %0 \n\t"
"subq $8 , %1 \n\t"
"jnz 1b \n\t"
"3: \n\t"
"vbroadcastss (%6) , %%xmm0 \n\t" // value from alpha
"vbroadcastss 4(%6) , %%xmm1 \n\t" // value from alpha
#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )
"vpermilps $0xb1 , %%ymm9 , %%ymm9 \n\t"
"vpermilps $0xb1 , %%ymm11, %%ymm11 \n\t"
"vaddsubps %%ymm9 , %%ymm8, %%ymm8 \n\t"
"vaddsubps %%ymm11, %%ymm10, %%ymm10 \n\t"
#else
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
"vpermilps $0xb1 , %%ymm10, %%ymm10 \n\t"
"vaddsubps %%ymm8 , %%ymm9 , %%ymm8 \n\t"
"vaddsubps %%ymm10, %%ymm11, %%ymm10 \n\t"
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
"vpermilps $0xb1 , %%ymm10, %%ymm10 \n\t"
#endif
"vmovsd (%3), %%xmm4 \n\t" // read y
"vmovsd 8(%3), %%xmm5 \n\t"
"vextractf128 $1, %%ymm8 , %%xmm9 \n\t"
"vextractf128 $1, %%ymm10, %%xmm11 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vshufpd $0x1, %%xmm8 , %%xmm8 , %%xmm9 \n\t"
"vshufpd $0x1, %%xmm10, %%xmm10, %%xmm11 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vmulps %%xmm8 , %%xmm1 , %%xmm9 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm8 , %%xmm0 , %%xmm8 \n\t" // t_r * alpha_r , t_i * alpha_r
"vmulps %%xmm10, %%xmm1 , %%xmm11 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm10, %%xmm0 , %%xmm10 \n\t" // t_r * alpha_r , t_i * alpha_r
#if !defined(XCONJ)
"vpermilps $0xb1 , %%xmm9 , %%xmm9 \n\t"
"vpermilps $0xb1 , %%xmm11, %%xmm11 \n\t"
"vaddsubps %%xmm9 , %%xmm8, %%xmm8 \n\t"
"vaddsubps %%xmm11, %%xmm10, %%xmm10 \n\t"
#else
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
"vpermilps $0xb1 , %%xmm10, %%xmm10 \n\t"
"vaddsubps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vaddsubps %%xmm10, %%xmm11, %%xmm10 \n\t"
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
"vpermilps $0xb1 , %%xmm10, %%xmm10 \n\t"
#endif
"vaddps %%xmm8 , %%xmm4 , %%xmm8 \n\t"
"vaddps %%xmm10, %%xmm5 , %%xmm10 \n\t"
"vmovsd %%xmm8 , (%3) \n\t"
"vmovsd %%xmm10, 8(%3) \n\t"
"vzeroupper \n\t"
:
"+r" (i), // 0
"+r" (n) // 1
:
"r" (x), // 2
"r" (y), // 3
"r" (ap[0]), // 4
"r" (ap[1]), // 5
"r" (alpha) // 6
: "cc",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);
}
#define HAVE_KERNEL_4x1 1
static void cgemv_kernel_4x1( BLASLONG n, FLOAT *ap, FLOAT *x, FLOAT *y, FLOAT *alpha) __attribute__ ((noinline));
static void cgemv_kernel_4x1( BLASLONG n, FLOAT *ap, FLOAT *x, FLOAT *y, FLOAT *alpha)
{
BLASLONG register i = 0;
__asm__ __volatile__
(
"vzeroupper \n\t"
"vxorps %%ymm8 , %%ymm8 , %%ymm8 \n\t" // temp
"vxorps %%ymm9 , %%ymm9 , %%ymm9 \n\t" // temp
"testq $0x04, %1 \n\t"
"jz 2f \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $8 , %0 \n\t"
"subq $4 , %1 \n\t"
"2: \n\t"
"cmpq $0, %1 \n\t"
"je 3f \n\t"
// ".align 16 \n\t"
"1: \n\t"
"prefetcht0 192(%4,%0,4) \n\t"
"vmovups (%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"prefetcht0 192(%2,%0,4) \n\t"
"vmovups (%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"vmovups 32(%4,%0,4), %%ymm4 \n\t" // 4 complex values from a0
"vmovups 32(%2,%0,4) , %%ymm6 \n\t" // 4 complex values from x
"vpermilps $0xb1, %%ymm6, %%ymm7 \n\t" // exchange real and imap parts
"vblendps $0x55, %%ymm6, %%ymm7, %%ymm0 \n\t" // only the real parts
"vblendps $0x55, %%ymm7, %%ymm6, %%ymm1 \n\t" // only the imag parts
"vfmadd231ps %%ymm4 , %%ymm0, %%ymm8 \n\t" // ar0*xr0,al0*xr0,ar1*xr1,al1*xr1
"vfmadd231ps %%ymm4 , %%ymm1, %%ymm9 \n\t" // ar0*xl0,al0*xl0,ar1*xl1,al1*xl1
"addq $16 , %0 \n\t"
"subq $8 , %1 \n\t"
"jnz 1b \n\t"
"3: \n\t"
"vbroadcastss (%5) , %%xmm0 \n\t" // value from alpha
"vbroadcastss 4(%5) , %%xmm1 \n\t" // value from alpha
#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )
"vpermilps $0xb1 , %%ymm9 , %%ymm9 \n\t"
"vaddsubps %%ymm9 , %%ymm8, %%ymm8 \n\t"
#else
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
"vaddsubps %%ymm8 , %%ymm9 , %%ymm8 \n\t"
"vpermilps $0xb1 , %%ymm8 , %%ymm8 \n\t"
#endif
"vmovsd (%3), %%xmm4 \n\t" // read y
"vextractf128 $1, %%ymm8 , %%xmm9 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vshufpd $0x1, %%xmm8 , %%xmm8 , %%xmm9 \n\t"
"vaddps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vmulps %%xmm8 , %%xmm1 , %%xmm9 \n\t" // t_r * alpha_i , t_i * alpha_i
"vmulps %%xmm8 , %%xmm0 , %%xmm8 \n\t" // t_r * alpha_r , t_i * alpha_r
#if !defined(XCONJ)
"vpermilps $0xb1 , %%xmm9 , %%xmm9 \n\t"
"vaddsubps %%xmm9 , %%xmm8, %%xmm8 \n\t"
#else
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
"vaddsubps %%xmm8 , %%xmm9 , %%xmm8 \n\t"
"vpermilps $0xb1 , %%xmm8 , %%xmm8 \n\t"
#endif
"vaddps %%xmm8 , %%xmm4 , %%xmm8 \n\t"
"vmovsd %%xmm8 , (%3) \n\t"
"vzeroupper \n\t"
:
"+r" (i), // 0
"+r" (n) // 1
:
"r" (x), // 2
"r" (y), // 3
"r" (ap), // 4
"r" (alpha) // 5
: "cc",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);
}