complex_128bit_256bit_512bit_BLOCK_template.c 137 KB
Newer Older
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
//    This file is part of ELPA.
//
//    The ELPA library was originally created by the ELPA consortium,
//    consisting of the following organizations:
//
//    - Max Planck Computing and Data Facility (MPCDF), formerly known as
//      Rechenzentrum Garching der Max-Planck-Gesellschaft (RZG),
//    - Bergische Universität Wuppertal, Lehrstuhl für angewandte
//      Informatik,
//    - Technische Universität München, Lehrstuhl für Informatik mit
//      Schwerpunkt Wissenschaftliches Rechnen ,
//    - Fritz-Haber-Institut, Berlin, Abt. Theorie,
//    - Max-Plack-Institut für Mathematik in den Naturwissenschaften,
//      Leipzig, Abt. Komplexe Strukutren in Biologie und Kognition,
//      and
//    - IBM Deutschland GmbH
//
//
//    This particular source code file contains additions, changes and
//    enhancements authored by Intel Corporation which is not part of
//    the ELPA consortium.
//
//    More information can be found here:
//    http://elpa.mpcdf.mpg.de/
//
//    ELPA is free software: you can redistribute it and/or modify
//    it under the terms of the version 3 of the license of the
//    GNU Lesser General Public License as published by the Free
//    Software Foundation.
//
//    ELPA is distributed in the hope that it will be useful,
//    but WITHOUT ANY WARRANTY; without even the implied warranty of
//    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
//    GNU Lesser General Public License for more details.
//
//    You should have received a copy of the GNU Lesser General Public License
//    along with ELPA. If not, see <http://www.gnu.org/licenses/>
//
//    ELPA reflects a substantial effort on the part of the original
//    ELPA consortium, and we ask you to respect the spirit of the
//    license that we chose: i.e., please contribute any changes you
//    may have back to the original ELPA library distribution, and keep
//    any derivatives of ELPA under the same license that we chose for
//    the original distribution, the GNU Lesser General Public License.
//
// Author: Andreas Marek, MPCDF, based on the double precision case of A. Heinecke
//
#include "config-f90.h"

#define CONCAT_8ARGS(a, b, c, d, e, f, g, h) CONCAT2_8ARGS(a, b, c, d, e, f, g, h)
#define CONCAT2_8ARGS(a, b, c, d, e, f, g, h) a ## b ## c ## d ## e ## f ## g ## h

#define CONCAT_7ARGS(a, b, c, d, e, f, g) CONCAT2_7ARGS(a, b, c, d, e, f, g)
#define CONCAT2_7ARGS(a, b, c, d, e, f, g) a ## b ## c ## d ## e ## f ## g

#define CONCAT_6ARGS(a, b, c, d, e, f) CONCAT2_6ARGS(a, b, c, d, e, f)
#define CONCAT2_6ARGS(a, b, c, d, e, f) a ## b ## c ## d ## e ## f

#define CONCAT_5ARGS(a, b, c, d, e) CONCAT2_5ARGS(a, b, c, d, e)
#define CONCAT2_5ARGS(a, b, c, d, e) a ## b ## c ## d ## e

#define CONCAT_4ARGS(a, b, c, d) CONCAT2_4ARGS(a, b, c, d)
#define CONCAT2_4ARGS(a, b, c, d) a ## b ## c ## d

#define CONCAT_3ARGS(a, b, c) CONCAT2_3ARGS(a, b, c)
#define CONCAT2_3ARGS(a, b, c) a ## b ## c

//define instruction set numbers
#define SSE_128 128
70
#define AVX_256 256
71
72
#define NEON_ARCH64_128 1285

73
#if VEC_SET == SSE_128 || VEC_SET == AVX_256 || VEC_SET == 512
74
75
76
77
#include <x86intrin.h>
#ifdef BLOCK2
#include <pmmintrin.h>
#endif
78
79
80

#define __forceinline __attribute__((always_inline))

81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
#endif


#include <complex.h>

#include <stdio.h>
#include <stdlib.h>

#ifdef BLOCK2
#define PREFIX double
#define BLOCK 2
#endif

#ifdef BLOCK1
#define PREFIX single
#define BLOCK 1
#endif

#if VEC_SET == SSE_128
#define SIMD_SET SSE
#endif

103
104
105
106
107
108
#if VEC_SET == AVX_256
#define SIMD_SET AVX_AVX2
#endif

#if VEC_SET == SSE_128

109
110
111
112
113
114
#ifdef DOUBLE_PRECISION_COMPLEX
#define offset 2
#define __SIMD_DATATYPE __m128d
#define _SIMD_LOAD _mm_load_pd
#define _SIMD_LOADU _mm_loadu_pd
#define _SIMD_STORE _mm_store_pd
115
#define _SIMD_STOREU _mm_storeu_pd
116
117
118
119
120
121
122
123
124
125
126
127
128
129
#define _SIMD_MUL _mm_mul_pd
#define _SIMD_ADD _mm_add_pd
#define _SIMD_XOR _mm_xor_pd
#define _SIMD_MADDSUB _mm_maddsub_pd
#define _SIMD_ADDSUB _mm_addsub_pd
#define _SIMD_SHUFFLE _mm_shuffle_pd
#define _SHUFFLE _MM_SHUFFLE2(0,1)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define offset 4
#define __SIMD_DATATYPE __m128
#define _SIMD_LOAD _mm_load_ps
#define _SIMD_LOADU _mm_loadu_ps
#define _SIMD_STORE _mm_store_ps
130
#define _SIMD_STOREU _mm_storeu_ps
131
132
133
134
135
136
137
138
139
#define _SIMD_MUL _mm_mul_ps
#define _SIMD_ADD _mm_add_ps
#define _SIMD_XOR _mm_xor_ps
#define _SIMD_MADDSUB _mm_maddsub_ps
#define _SIMD_ADDSUB _mm_addsub_ps
#define _SIMD_SHUFFLE _mm_shuffle_ps
#define _SHUFFLE 0xb1
#endif

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
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256

#ifdef DOUBLE_PRECISION_COMPLEX
#define offset 4
#define __SIMD_DATATYPE __m256d
#define _SIMD_LOAD _mm256_load_pd
#define _SIMD_LOADU 1
#define _SIMD_STORE _mm256_store_pd
#define _SIMD_STOREU 1
#define _SIMD_MUL _mm256_mul_pd
#define _SIMD_ADD _mm256_add_pd
#define _SIMD_XOR _mm256_xor_pd
#define _SIMD_BROADCAST _mm256_broadcast_sd
#define _SIMD_MADDSUB 1
#define _SIMD_ADDSUB _mm256_addsub_pd
#define _SIMD_SHUFFLE _mm256_shuffle_pd
#define _SHUFFLE 0x5
#ifdef HAVE_AVX2

#ifdef __FMA4__
#define __ELPA_USE_FMA__
#define _mm256_FMADDSUB_pd(a,b,c) _mm256_maddsub_pd(a,b,c)
#define _mm256_FMSUBADD_pd(a,b,c) _mm256_msubadd_pd(a,b,c)
#endif

#ifdef __AVX2__
#define __ELPA_USE_FMA__
#define _mm256_FMADDSUB_pd(a,b,c) _mm256_fmaddsub_pd(a,b,c)
#define _mm256_FMSUBADD_pd(a,b,c) _mm256_fmsubadd_pd(a,b,c)
#endif

#endif /* HAVE_AVX2 */

#define _SIMD_FMADDSUB _mm256_FMADDSUB_pd
#define _SIMD_FMSUBADD _mm256_FMSUBADD_pd

#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#define offset 8
#define __SIMD_DATATYPE __m256
#define _SIMD_LOAD _mm256_load_ps
#define _SIMD_LOADU 1
#define _SIMD_STORE _mm256_store_ps
#define _SIMD_STOREU 1
#define _SIMD_MUL _mm256_mul_ps
#define _SIMD_ADD _mm256_add_ps
#define _SIMD_XOR _mm256_xor_ps
#define _SIMD_BROADCAST  _mm256_broadcast_ss
#define _SIMD_MADDSUB 1
#define _SIMD_ADDSUB _mm256_addsub_ps
#define _SIMD_SHUFFLE _mm_shuffle_ps
#define _SHUFFLE 0xb1

#ifdef HAVE_AVX2

#ifdef __FMA4__
#define __ELPA_USE_FMA__
#define _mm256_FMADDSUB_ps(a,b,c) _mm256_maddsub_ps(a,b,c)
#define _mm256_FMSUBADD_ps(a,b,c) _mm256_msubadd_ps(a,b,c)
#endif

#ifdef __AVX2__
#define __ELPA_USE_FMA__
#define _mm256_FMADDSUB_ps(a,b,c) _mm256_fmaddsub_ps(a,b,c)
#define _mm256_FMSUBADD_ps(a,b,c) _mm256_fmsubadd_ps(a,b,c)
#endif

#endif /* HAVE_AVX2 */

#define _SIMD_FMADDSUB _mm256_FMADDSUB_ps
#define _SIMD_FMSUBADD _mm256_FMSUBADD_ps

#endif /* SINGLE_PRECISION_COMPLEX */

#endif /* VEC_SET == AVX_256 */



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
#define __forceinline __attribute__((always_inline))

#ifdef HAVE_SSE_INTRINSICS
#undef __AVX__
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
#define WORD_LENGTH double
#define DATA_TYPE double complex
#define DATA_TYPE_PTR double complex*
#define DATA_TYPE_REAL double
#define DATA_TYPE_REAL_PTR double*
#endif

#ifdef SINGLE_PRECISION_COMPLEX
#define WORD_LENGTH single
#define DATA_TYPE float complex
#define DATA_TYPE_PTR float complex*
#define DATA_TYPE_REAL float
#define DATA_TYPE_REAL_PTR float*
#endif


#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 6
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 12
#endif
#endif /* VEC_SET  == SSE_128 */

255
256
257
258
259
260
261
262
263
264
265
#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 12
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 24
#endif
#endif /* VEC_SET  == AVX_256 */

266
267
268
269
270
271
272
273
274
//Forward declaration
static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq 
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif

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
#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 5
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 10
#endif
#endif /* VEC_SET  == SSE_128 */


#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 10
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 20
#endif
#endif /* VEC_SET  == AVX_256 */


static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif


308
309
310
311
312
313
314
315
316
317
#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 4
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 8
#endif
#endif /* VEC_SET  == SSE_128 */
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


#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 8
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 16
#endif
#endif /* VEC_SET  == AVX_256 */


static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif

#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 3
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 6
#endif
#endif /* VEC_SET  == SSE_128 */


#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 6
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 12
#endif
#endif /* VEC_SET  == AVX_256 */


364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif

#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 4
#endif
#endif /* VEC_SET  == SSE_128 */
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

#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 4
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 8
#endif
#endif /* VEC_SET  == AVX_256 */


static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif

#if VEC_SET  == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 1
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 2
#endif
#endif /* VEC_SET  == SSE_128 */

#if VEC_SET  == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#undef ROW_LENGTH 
#define ROW_LENGTH 4
#endif
#endif /* VEC_SET  == AVX_256 */

425
426
427
428
429
430
431
432
static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH)(DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		                       );
#endif
#ifdef BLOCK2
                                       ,int ldh, DATA_TYPE s);
#endif

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
/*
!f>#ifdef HAVE_SSE_INTRINSICS
!f> interface
!f>   subroutine single_hh_trafo_complex_SSE_1hv_double(q, hh, pnb, pnq, pldq) &
!f>                             bind(C, name="single_hh_trafo_complex_SSE_1hv_double")
!f>     use, intrinsic :: iso_c_binding
!f>     integer(kind=c_int)     :: pnb, pnq, pldq
!f>     ! complex(kind=c_double_complex)     :: q(*)
!f>     type(c_ptr), value                   :: q
!f>     complex(kind=c_double_complex)     :: hh(pnb,2)
!f>   end subroutine
!f> end interface
!f>#endif
*/

/*
!f>#ifdef HAVE_SSE_INTRINSICS
!f> interface
!f>   subroutine single_hh_trafo_complex_SSE_1hv_single(q, hh, pnb, pnq, pldq) &
!f>                             bind(C, name="single_hh_trafo_complex_SSE_1hv_single")
!f>     use, intrinsic :: iso_c_binding
!f>     integer(kind=c_int)     :: pnb, pnq, pldq
!f>     ! complex(kind=c_float_complex)   :: q(*)
!f>     type(c_ptr), value                :: q
!f>     complex(kind=c_float_complex)   :: hh(pnb,2)
!f>   end subroutine
!f> end interface
!f>#endif
*/

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
/*
!f>#ifdef HAVE_SSE_INTRINSICS
!f> interface
!f>   subroutine double_hh_trafo_complex_SSE_2hv_double(q, hh, pnb, pnq, pldq, pldh) &
!f>                             bind(C, name="double_hh_trafo_complex_SSE_2hv_double")
!f>     use, intrinsic :: iso_c_binding
!f>     integer(kind=c_int)     :: pnb, pnq, pldq, pldh
!f>     ! complex(kind=c_double_complex)     :: q(*)
!f>     type(c_ptr), value                   :: q
!f>     complex(kind=c_double_complex)     :: hh(pnb,2)
!f>   end subroutine
!f> end interface
!f>#endif
*/

/*
!f>#ifdef HAVE_SSE_INTRINSICS
!f> interface
!f>   subroutine double_hh_trafo_complex_SSE_2hv_single(q, hh, pnb, pnq, pldq, pldh) &
!f>                             bind(C, name="double_hh_trafo_complex_SSE_2hv_single")
!f>     use, intrinsic :: iso_c_binding
!f>     integer(kind=c_int)     :: pnb, pnq, pldq, pldh
!f>     ! complex(kind=c_float_complex)   :: q(*)
!f>     type(c_ptr), value                :: q
!f>     complex(kind=c_float_complex)   :: hh(pnb,2)
!f>   end subroutine
!f> end interface
!f>#endif
*/


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
void CONCAT_7ARGS(PREFIX,_hh_trafo_complex_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int* pnb, int* pnq, int* pldq
#ifdef BLOCK1
		  )
#endif
#ifdef BLOCK2
                  ,int* pldh)
#endif
{

     int i, worked_on;
     int nb = *pnb;
     int nq = *pldq;
     int ldq = *pldq;
#ifdef BLOCK2
     int ldh = *pldh;

     DATA_TYPE s = conj(hh[(ldh)+1])*1.0;

     for (i = 2; i < nb; i++)
     {
             s += hh[i-1] * conj(hh[(i+ldh)]);
     }
#endif

     worked_on = 0;
521

522
523
#ifdef BLOCK1

524
#if VEC_SET == SSE_128
525
526
527
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 6
#define STEP_SIZE 6
528
#define UPPER_BOUND 5
529
530
531
532
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 12
#define STEP_SIZE 12
533
534
535
536
537
538
539
540
541
542
543
544
545
546
#define UPPER_BOUND 10
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 12
#define STEP_SIZE 12
#define UPPER_BOUND 10
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 24
#define STEP_SIZE 24
#define UPPER_BOUND 20
547
#endif
548
549
#endif /* VEC_SET == AVX_256 */

550
551
552
553
554
555
556
557
558
559
        for (i = 0; i < nq - UPPER_BOUND; i+= STEP_SIZE)
        {

            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }
        if (nq == i) {
          return;
        }

560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
#if VEC_SET == SSE_128
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 5
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 10
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 10
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 20
#endif
#endif /* VEC_SET == AVX_256 */

        if (nq-i == ROW_LENGTH)
        {
            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }


#if VEC_SET == SSE_128
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 8
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 8
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 16
#endif
#endif /* VEC_SET == AVX_256 */

        if (nq-i == ROW_LENGTH)
        {
            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }

#if VEC_SET == SSE_128
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 3
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 6
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 6
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 12
#endif
#endif /* VEC_SET == AVX_256 */

        if (nq-i == ROW_LENGTH)
        {
            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }

#if VEC_SET == SSE_128
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
650
651
652
653
654
655
656
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 8
#endif
657
658
659
#endif /* VEC_SET == AVX_256 */


660
661
662
663
664
665
        if (nq-i == ROW_LENGTH)
        {
            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }

666
667
668
669
670
671
672
673
674
675
676
#if VEC_SET == SSE_128
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 1
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 2
#endif
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
677
678
679
680
681
682
683
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#endif
684
#endif /* VEC_SET == AVX_256 */
685
686
687
688
689
        if (nq-i == ROW_LENGTH)
        {
            CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq);
	    worked_on += ROW_LENGTH;
        }
690

691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
#endif /* BLOCK1 */

#ifdef BLOCK2

#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#define STEP_SIZE 4
#define UPPER_BOUND 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 8
#define STEP_SIZE 8
#define UPPER_BOUND 4
#endif

    for (i = 0; i < nq - UPPER_BOUND; i+=STEP_SIZE)
    {
         CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq, ldh, s);
	 worked_on +=ROW_LENGTH;
    }
 
    if (nq == i)
    {
      return;
    }
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 2
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 4
#endif
    if (nq-i == ROW_LENGTH)
    {
        CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq, ldh, s);
        worked_on += ROW_LENGTH;
    }

730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
#undef ROW_LENGTH
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 1
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 2
#endif
    if (nq-i == ROW_LENGTH)
    {
        CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (&q[i], hh, nb, ldq, ldh, s);
        worked_on += ROW_LENGTH;
    }
#endif /* BLOCK2 */

//#ifdef WITH_DEBUG
    if (worked_on != nq)
    {
      printf("Error in complex SIMD_SET BLOCK BLOCK kernel %d %d\n", worked_on, nq);
      abort();
    }
//#endif
751
752
753
754


}

755
#if VEC_SET == SSE_128
756
757
758
759
760
761
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 6
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 12
#endif
762
763
764
765
766
767
768
769
770
771
772
#endif /* VEC_SET == SSE_128 */

#if VEC_SET == AVX_256
#ifdef DOUBLE_PRECISION_COMPLEX
#define ROW_LENGTH 12
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define ROW_LENGTH 24
#endif
#endif /* VEC_SET == AVX_256 */

773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
static __forceinline void CONCAT_8ARGS(hh_trafo_complex_kernel_,ROW_LENGTH,_,SIMD_SET,_,BLOCK,hv_,WORD_LENGTH) (DATA_TYPE_PTR q, DATA_TYPE_PTR hh, int nb, int ldq
#ifdef BLOCK1
		)
#endif
#ifdef BLOCK2
                ,int ldh, DATA_TYPE s)
#endif
{

    DATA_TYPE_REAL_PTR q_dbl = (DATA_TYPE_REAL_PTR)q;
    DATA_TYPE_REAL_PTR hh_dbl = (DATA_TYPE_REAL_PTR)hh;
#ifdef BLOCK2
    DATA_TYPE_REAL_PTR s_dbl = (DATA_TYPE_REAL_PTR)(&s);
#endif

    __SIMD_DATATYPE x1, x2, x3, x4, x5, x6;
    __SIMD_DATATYPE q1, q2, q3, q4, q5, q6;
#ifdef BLOCK2
    __SIMD_DATATYPE y1, y2, y3, y4, y5, y6;
    __SIMD_DATATYPE h2_real, h2_imag;
#endif
    __SIMD_DATATYPE h1_real, h1_imag;
    __SIMD_DATATYPE tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
    int i=0;

#if VEC_SET == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
    __SIMD_DATATYPE sign = (__SIMD_DATATYPE)_mm_set_epi64x(0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
    __SIMD_DATATYPE sign = (__SIMD_DATATYPE)_mm_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000);
#endif
#endif /* VEC_SET == SSE_128 */

#ifdef BLOCK2
     x1 = _SIMD_LOAD(&q_dbl[(2*ldq)+0]);
     x2 = _SIMD_LOAD(&q_dbl[(2*ldq)+offset]);
     x3 = _SIMD_LOAD(&q_dbl[(2*ldq)+2*offset]);
     x4 = _SIMD_LOAD(&q_dbl[(2*ldq)+3*offset]);
     x5 = _SIMD_LOAD(&q_dbl[(2*ldq)+4*offset]);
     x6 = _SIMD_LOAD(&q_dbl[(2*ldq)+5*offset]);


#if VEC_SET == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
     h2_real = _mm_loaddup_pd(&hh_dbl[(ldh+1)*2]);
     h2_imag = _mm_loaddup_pd(&hh_dbl[((ldh+1)*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
     h2_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(ldh+1)*2]) )));
     h2_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[((ldh+1)*2)+1]) )));
#endif
#endif /* VEC_SET == SSE_128 */

#ifndef __ELPA_USE_FMA__
     // conjugate
     h2_imag = _SIMD_XOR(h2_imag, sign);
#endif

     y1 = _SIMD_LOAD(&q_dbl[0]);
     y2 = _SIMD_LOAD(&q_dbl[offset]);
     y3 = _SIMD_LOAD(&q_dbl[2*offset]);
     y4 = _SIMD_LOAD(&q_dbl[3*offset]);
     y5 = _SIMD_LOAD(&q_dbl[4*offset]);
     y6 = _SIMD_LOAD(&q_dbl[5*offset]);

     tmp1 = _SIMD_MUL(h2_imag, x1);
#ifdef __ELPA_USE_FMA__
     y1 = _SIMD_ADD(y1, _mm_msubadd_pd(h2_real, x1, _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#else
     y1 = _SIMD_ADD(y1, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x1), _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#endif
     tmp2 = _SIMD_MUL(h2_imag, x2);
#ifdef __ELPA_USE_FMA__
     y2 = _SIMD_ADD(y2, _mm_msubadd_pd(h2_real, x2, _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#else
     y2 = _SIMD_ADD(y2, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x2), _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#endif

     tmp3 = _SIMD_MUL(h2_imag, x3);
#ifdef __ELPA_USE_FMA__
     y3 = _SIMD_ADD(y3, _mm_msubadd_pd(h2_real, x3, _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#else
     y3 = _SIMD_ADD(y3, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x3), _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#endif
     tmp4 = _SIMD_MUL(h2_imag, x4);
#ifdef __ELPA_USE_FMA__
     y4 = _SIMD_ADD(y4, _mm_msubadd_pd(h2_real, x4, _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#else
     y4 = _SIMD_ADD(y4, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x4), _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#endif

     tmp5 = _SIMD_MUL(h2_imag, x5);
#ifdef __ELPA_USE_FMA__
     y5 = _SIMD_ADD(y5, _mm_msubadd_pd(h2_real, x5, _SIMD_SHUFFLE(tmp5, tmp5, _SHUFFLE)));
#else
     y5 = _SIMD_ADD(y5, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x5), _SIMD_SHUFFLE(tmp5, tmp5, _SHUFFLE)));
#endif
     tmp6 = _SIMD_MUL(h2_imag, x6);
#ifdef __ELPA_USE_FMA__
     y6 = _SIMD_ADD(y6, _mm_msubadd_pd(h2_real, x6, _SIMD_SHUFFLE(tmp6, tmp6, _SHUFFLE)));
#else
     y6 = _SIMD_ADD(y6, _SIMD_ADDSUB( _SIMD_MUL(h2_real, x6), _SIMD_SHUFFLE(tmp6, tmp6, _SHUFFLE)));
#endif

#endif /* BLOCK2 */

#ifdef BLOCK1
    x1 = _SIMD_LOAD(&q_dbl[0]);
    x2 = _SIMD_LOAD(&q_dbl[offset]);
    x3 = _SIMD_LOAD(&q_dbl[2*offset]);
    x4 = _SIMD_LOAD(&q_dbl[3*offset]);
    x5 = _SIMD_LOAD(&q_dbl[4*offset]);
    x6 = _SIMD_LOAD(&q_dbl[5*offset]);
#endif

    for (i = BLOCK; i < nb; i++)
    {

#if VEC_SET == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
        h1_real = _mm_loaddup_pd(&hh_dbl[(i-BLOCK+1)*2]);
        h1_imag = _mm_loaddup_pd(&hh_dbl[((i-BLOCK+1)*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
        h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i-BLOCK+1)*2]) )));
        h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[((i-BLOCK+1)*2)+1]) )));
#endif
#endif /* VEC_SET == SSE_128 */

#ifndef __ELPA_USE_FMA__
        // conjugate
        h1_imag = _SIMD_XOR(h1_imag, sign);
#endif

        q1 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+0]);
        q2 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+offset]);
        q3 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
        q4 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
        q5 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+4*offset]);
        q6 = _SIMD_LOAD(&q_dbl[(2*i*ldq)+5*offset]);

        tmp1 = _SIMD_MUL(h1_imag, q1);

#ifdef __ELPA_USE_FMA__
        x1 = _SIMD_ADD(x1, _mm_msubadd_pd(h1_real, q1, _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#else
        x1 = _SIMD_ADD(x1, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q1), _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#endif
        tmp2 = _SIMD_MUL(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
        x2 = _SIMD_ADD(x2, _mm_msubadd_pd(h1_real, q2, _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#else
        x2 = _SIMD_ADD(x2, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q2), _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#endif
        tmp3 = _SIMD_MUL(h1_imag, q3);
#ifdef __ELPA_USE_FMA__
        x3 = _SIMD_ADD(x3, _mm_msubadd_pd(h1_real, q3, _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#else
        x3 = _SIMD_ADD(x3, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q3), _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#endif

        tmp4 = _SIMD_MUL(h1_imag, q4);
#ifdef __ELPA_USE_FMA__
        x4 = _SIMD_ADD(x4, _mm_msubadd_pd(h1_real, q4, _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#else
        x4 = _SIMD_ADD(x4, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q4), _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#endif
        tmp5 = _SIMD_MUL(h1_imag, q5);
#ifdef __ELPA_USE_FMA__
        x5 = _SIMD_ADD(x5, _mm_msubadd_pd(h1_real, q5, _SIMD_SHUFFLE(tmp5, tmp5, _SHUFFLE)));
#else
        x5 = _SIMD_ADD(x5, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q5), _SIMD_SHUFFLE(tmp5, tmp5, _SHUFFLE)));
#endif
        tmp6 = _SIMD_MUL(h1_imag, q6);
#ifdef __ELPA_USE_FMA__
        x6 = _SIMD_ADD(x6, _mm_msubadd_pd(h1_real, q6, _SIMD_SHUFFLE(tmp6, tmp6, _SHUFFLE)));
#else
        x6 = _SIMD_ADD(x6, _SIMD_ADDSUB( _SIMD_MUL(h1_real, q6), _SIMD_SHUFFLE(tmp6, tmp6, _SHUFFLE)));
#endif

#ifdef BLOCK2

#if VEC_SET == SSE_128
#ifdef DOUBLE_PRECISION_COMPLEX
          h2_real = _mm_loaddup_pd(&hh_dbl[(ldh+i)*2]);
          h2_imag = _mm_loaddup_pd(&hh_dbl[((ldh+i)*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
          h2_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(ldh+i)*2]) )));
          h2_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[((ldh+i)*2)+1]) )));
#endif
#endif /* VEC_SET == SSE_128 */

#ifndef __ELPA_USE_FMA__
          // conjugate
          h2_imag = _SIMD_XOR(h2_imag, sign);
#endif

          tmp1 = _SIMD_MUL(h2_imag, q1);
#ifdef __ELPA_USE_FMA__
          y1 = _SIMD_ADD(y1, _mm_msubadd_pd(h2_real, q1, _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#else
          y1 = _SIMD_ADD(y1, _SIMD_ADDSUB( _SIMD_MUL(h2_real, q1), _SIMD_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
#endif
          tmp2 = _SIMD_MUL(h2_imag, q2);
#ifdef __ELPA_USE_FMA__
          y2 = _SIMD_ADD(y2, _mm_msubadd_pd(h2_real, q2, _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#else
          y2 = _SIMD_ADD(y2, _SIMD_ADDSUB( _SIMD_MUL(h2_real, q2), _SIMD_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
#endif

          tmp3 = _SIMD_MUL(h2_imag, q3);
#ifdef __ELPA_USE_FMA__
          y3 = _SIMD_ADD(y3, _mm_msubadd_pd(h2_real, q3, _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#else
          y3 = _SIMD_ADD(y3, _SIMD_ADDSUB( _SIMD_MUL(h2_real, q3), _SIMD_SHUFFLE(tmp3, tmp3, _SHUFFLE)));
#endif
          tmp4 = _SIMD_MUL(h2_imag, q4);
#ifdef __ELPA_USE_FMA__
          y4 = _SIMD_ADD(y4, _mm_msubadd_pd(h2_real, q4, _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#else
          y4 = _SIMD_ADD(y4, _SIMD_ADDSUB( _SIMD_MUL(h2_real, q4), _SIMD_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
#endif

          tmp5 = _SIMD_MUL(h2_imag, q5);
#ifdef __ELPA_USE_FMA__
          y5 = _SIMD_ADD(y5, _mm_msubadd_pd(h2_real, q5, _SIMD_SHUFFLE(tmp5, tmp5, _SHUFFLE)));