elpa2_kernels_complex_sse_1hv_template.Xc 31.8 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
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
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
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
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
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
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
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
//    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.
//
//
// --------------------------------------------------------------------------------------------------
//
// This file contains the compute intensive kernels for the Householder transformations.
// It should be compiled with the highest possible optimization level.
//
// On Intel Nehalem or Intel Westmere or AMD Magny Cours use -O3 -msse3
// On Intel Sandy Bridge use -O3 -mavx
//
// Copyright of the original code rests with the authors inside the ELPA
// consortium. The copyright of any additional modifications shall rest
// with their original authors, but shall adhere to the licensing terms
// distributed along with the original code in the file "COPYING".
//
// Author: Alexander Heinecke (alexander.heinecke@mytum.de)
// Adapted for building a shared-library by Andreas Marek, MPCDF (andreas.marek@mpcdf.mpg.de)
// --------------------------------------------------------------------------------------------------

#include "config-f90.h"

#include <complex.h>
#include <x86intrin.h>

#ifdef DOUBLE_PRECISION_COMPLEX
#define offset 2
#define __SSE_DATATYPE __m128d
#define _SSE_LOAD _mm_load_pd
#define _SSE_STORE _mm_store_pd
#define _SSE_MUL _mm_mul_pd
#define _SSE_ADD _mm_add_pd
#define _SSE_XOR _mm_xor_pd
#define _SSE_MADDSUB _mm_maddsub_pd
#define _SSE_ADDSUB _mm_addsub_pd
#define _SSE_SHUFFLE _mm_shuffle_pd
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define offset 4
#define __SSE_DATATYPE __m128
#define _SSE_LOAD _mm_load_ps
#define _SSE_STORE _mm_store_ps
#define _SSE_MUL _mm_mul_ps
#define _SSE_ADD _mm_add_ps
#define _SSE_XOR _mm_xor_ps
#define _SSE_MADDSUB _mm_maddsub_ps
#define _SSE_ADDSUB _mm_addsub_ps
#define _SSE_SHUFFLE _mm_shuffle_ps
#endif

#define __forceinline __attribute__((always_inline))

#ifdef HAVE_SSE_INTRINSICS
#undef __AVX__
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
//Forward declaration
static __forceinline void hh_trafo_complex_kernel_6_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
static __forceinline void hh_trafo_complex_kernel_4_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
static __forceinline void hh_trafo_complex_kernel_2_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
#endif

#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_6_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
static __forceinline void hh_trafo_complex_kernel_4_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
static __forceinline void hh_trafo_complex_kernel_2_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
/*
!f>#ifdef WITH_COMPLEX_SSE_BLOCK1_KERNEL
!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
*/
#endif

#ifdef SINGLE_PRECISION_COMPLEX
/*
!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
*/
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
void single_hh_trafo_complex_sse_1hv_double(double complex* q, double complex* hh, int* pnb, int* pnq, int* pldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
void single_hh_trafo_complex_sse_1hv_single(float complex* q, float complex* hh, int* pnb, int* pnq, int* pldq)
#endif
{
	int i;
	int nb = *pnb;
	int nq = *pldq;
	int ldq = *pldq;
	//int ldh = *pldh;

	for (i = 0; i < nq-4; i+=6)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_6_SSE_1hv_double(&q[i], hh, nb, ldq);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_6_SSE_1hv_single(&q[i], hh, nb, ldq);
#endif
	}
	if (nq-i == 0) {
	  return;
	} else {

	if (nq-i > 2)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_4_SSE_1hv_double(&q[i], hh, nb, ldq);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_4_SSE_1hv_single(&q[i], hh, nb, ldq);
#endif
	}
	else
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_2_SSE_1hv_double(&q[i], hh, nb, ldq);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		hh_trafo_complex_kernel_2_SSE_1hv_single(&q[i], hh, nb, ldq);
#endif
	}
    }
}

#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_6_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_6_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq)
#endif
{

#ifdef DOUBLE_PRECISION_COMPLEX
	double* q_dbl = (double*)q;
	double* hh_dbl = (double*)hh;
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	float* q_dbl = (float*)q;
	float* hh_dbl = (float*)hh;
#endif
	__SSE_DATATYPE x1, x2, x3, x4, x5, x6;
	__SSE_DATATYPE q1, q2, q3, q4, q5, q6;
	__SSE_DATATYPE h1_real, h1_imag;
	__SSE_DATATYPE tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi64x(0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000);
#endif

	x1 = _SSE_LOAD(&q_dbl[0]);
	x2 = _SSE_LOAD(&q_dbl[offset]);
	x3 = _SSE_LOAD(&q_dbl[2*offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
	x4 = _SSE_LOAD(&q_dbl[3*offset]);
	x5 = _SSE_LOAD(&q_dbl[4*offset]);
	x6 = _SSE_LOAD(&q_dbl[5*offset]);
#endif
	for (i = 1; i < nb; i++)
	{

#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif
#ifndef __ELPA_USE_FMA__
		// conjugate
		h1_imag = _SSE_XOR(h1_imag, sign);
#endif

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

		tmp1 = _SSE_MUL(h1_imag, q1);

#ifdef DOUBLE_PRECISION_COMPLEX

#ifdef __ELPA_USE_FMA__
		x1 = _SSE_ADD(x1, _mm_msubadd_pd(h1_real, q1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		x1 = _SSE_ADD(x1, _SSE_ADDSUB( _SSE_MUL(h1_real, q1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif
		tmp2 = _SSE_MUL(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
		x2 = _SSE_ADD(x2, _mm_msubadd_pd(h1_real, q2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		x2 = _SSE_ADD(x2, _SSE_ADDSUB( _SSE_MUL(h1_real, q2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
		tmp3 = _SSE_MUL(h1_imag, q3);
#ifdef __ELPA_USE_FMA__
		x3 = _SSE_ADD(x3, _mm_msubadd_pd(h1_real, q3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#else
		x3 = _SSE_ADD(x3, _SSE_ADDSUB( _SSE_MUL(h1_real, q3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#endif

#endif /* DOUBLE_PRECISION_COMPEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		x1 = _mm_add_ps(x1, _mm_msubadd_ps(h1_real, q1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		x1 = _mm_add_ps(x1, _mm_addsub_ps( _mm_mul_ps(h1_real, q1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
		tmp2 = _mm_mul_ps(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
		x2 = _mm_add_ps(x2, _mm_msubadd_ps(h1_real, q2, _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#else
		x2 = _mm_add_ps(x2, _mm_addsub_ps( _mm_mul_ps(h1_real, q2), _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#endif
		tmp3 = _mm_mul_ps(h1_imag, q3);
#ifdef __ELPA_USE_FMA__
		x3 = _mm_add_ps(x3, _mm_msubadd_ps(h1_real, q3, _mm_shuffle_ps(tmp3, tmp3, 0xb1)));
#else
		x3 = _mm_add_ps(x3, _mm_addsub_ps( _mm_mul_ps(h1_real, q3), _mm_shuffle_ps(tmp3, tmp3, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */


#ifdef DOUBLE_PRECISION_COMPLEX
		tmp4 = _SSE_MUL(h1_imag, q4);
#ifdef __ELPA_USE_FMA__
		x4 = _SSE_ADD(x4, _mm_msubadd_pd(h1_real, q4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#else
		x4 = _SSE_ADD(x4, _SSE_ADDSUB( _SSE_MUL(h1_real, q4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#endif
		tmp5 = _SSE_MUL(h1_imag, q5);
#ifdef __ELPA_USE_FMA__
		x5 = _SSE_ADD(x5, _mm_msubadd_pd(h1_real, q5, _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1))));
#else
		x5 = _SSE_ADD(x5, _SSE_ADDSUB( _SSE_MUL(h1_real, q5), _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1))));
#endif
		tmp6 = _SSE_MUL(h1_imag, q6);
#ifdef __ELPA_USE_FMA__
		x6 = _SSE_ADD(x6, _mm_msubadd_pd(h1_real, q6, _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1))));
#else
		x6 = _SSE_ADD(x6, _SSE_ADDSUB( _SSE_MUL(h1_real, q6), _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1))));
#endif

#endif /* DOUBLE_PRECISION_COMPLEX */
	}

#ifdef DOUBLE_PRECISION_COMPLEX
	h1_real = _mm_loaddup_pd(&hh_dbl[0]);
	h1_imag = _mm_loaddup_pd(&hh_dbl[1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[0]) )));
	h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[1]) )));
#endif
	h1_real = _SSE_XOR(h1_real, sign);
	h1_imag = _SSE_XOR(h1_imag, sign);

	tmp1 = _SSE_MUL(h1_imag, x1);

#ifdef DOUBLE_PRECISION_COMPLEX

#ifdef __ELPA_USE_FMA__
	x1 = _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#else
	x1 = _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#endif
	tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
	x2 = _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#else
	x2 = _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#endif
	tmp3 = _SSE_MUL(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
	x3 = _SSE_MADDSUB(h1_real, x3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1)));
#else
	x3 = _SSE_ADDSUB( _SSE_MUL(h1_real, x3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
	x1 = _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#else
	x1 = _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#endif
	tmp2 = _mm_mul_ps(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
	x2 = _mm_maddsub_ps(h1_real, x2, _mm_shuffle_ps(tmp2, tmp2, 0xb1));
#else
	x2 = _mm_addsub_ps( _mm_mul_ps(h1_real, x2), _mm_shuffle_ps(tmp2, tmp2, 0xb1));
#endif
	tmp3 = _mm_mul_ps(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
	x3 = _mm_maddsub_ps(h1_real, x3, _mm_shuffle_ps(tmp3, tmp3, 0xb1));
#else
	x3 = _mm_addsub_ps( _mm_mul_ps(h1_real, x3), _mm_shuffle_ps(tmp3, tmp3, 0xb1));
#endif

#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
	tmp4 = _SSE_MUL(h1_imag, x4);
#ifdef __ELPA_USE_FMA__
	x4 = _SSE_MADDSUB(h1_real, x4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1)));
#else
	x4 = _SSE_ADDSUB( _SSE_MUL(h1_real, x4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1)));
#endif
	tmp5 = _SSE_MUL(h1_imag, x5);
#ifdef __ELPA_USE_FMA__
	x5 = _SSE_MADDSUB(h1_real, x5, _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1)));
#else
	x5 = _SSE_ADDSUB( _SSE_MUL(h1_real, x5), _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1)));
#endif
	tmp6 = _SSE_MUL(h1_imag, x6);
#ifdef __ELPA_USE_FMA__
	x6 = _SSE_MADDSUB(h1_real, x6, _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1)));
#else
	x6 = _SSE_ADDSUB( _SSE_MUL(h1_real, x6), _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

	q1 = _SSE_LOAD(&q_dbl[0]);
	q2 = _SSE_LOAD(&q_dbl[offset]);
	q3 = _SSE_LOAD(&q_dbl[2*offset]);
#ifdef DOUBLE_PRECISION_COMPLEX 
	q4 = _SSE_LOAD(&q_dbl[3*offset]);
	q5 = _SSE_LOAD(&q_dbl[4*offset]);
	q6 = _SSE_LOAD(&q_dbl[5*offset]);
#endif

	q1 = _SSE_ADD(q1, x1);
	q2 = _SSE_ADD(q2, x2);
	q3 = _SSE_ADD(q3, x3);
#ifdef DOUBLE_PRECISION_COMPLEX 
	q4 = _SSE_ADD(q4, x4);
	q5 = _SSE_ADD(q5, x5);
	q6 = _SSE_ADD(q6, x6);
#endif

	_SSE_STORE(&q_dbl[0], q1);
	_SSE_STORE(&q_dbl[offset], q2);
	_SSE_STORE(&q_dbl[2*offset], q3);
#ifdef DOUBLE_PRECISION_COMPLEX 
	_SSE_STORE(&q_dbl[3*offset], q4);
	_SSE_STORE(&q_dbl[4*offset], q5);
	_SSE_STORE(&q_dbl[5*offset], q6);
#endif
	for (i = 1; i < nb; i++)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif

		q1 = _SSE_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _SSE_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _SSE_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
		q4 = _SSE_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
		q5 = _SSE_LOAD(&q_dbl[(2*i*ldq)+4*offset]);
		q6 = _SSE_LOAD(&q_dbl[(2*i*ldq)+5*offset]);
#endif
		tmp1 = _SSE_MUL(h1_imag, x1);
#ifdef DOUBLE_PRECISION_COMPLEX

#ifdef __ELPA_USE_FMA__
		q1 = _SSE_ADD(q1, _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		q1 = _SSE_ADD(q1, _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif
		tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
		q2 = _SSE_ADD(q2, _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		q2 = _SSE_ADD(q2, _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
		tmp3 = _SSE_MUL(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
		q3 = _SSE_ADD(q3, _SSE_MADDSUB(h1_real, x3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#else
		q3 = _SSE_ADD(q3, _SSE_ADDSUB( _SSE_MUL(h1_real, x3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		q1 = _mm_add_ps(q1, _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		q1 = _mm_add_ps(q1, _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
		tmp2 = _mm_mul_ps(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
		q2 = _mm_add_ps(q2, _mm_maddsub_ps(h1_real, x2, _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#else
		q2 = _mm_add_ps(q2, _mm_addsub_ps( _mm_mul_ps(h1_real, x2), _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#endif
		tmp3 = _mm_mul_ps(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
		q3 = _mm_add_ps(q3, _mm_maddsub_ps(h1_real, x3, _mm_shuffle_ps(tmp3, tmp3, 0xb1)));
#else
		q3 = _mm_add_ps(q3, _mm_addsub_ps( _mm_mul_ps(h1_real, x3), _mm_shuffle_ps(tmp3, tmp3, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
		tmp4 = _SSE_MUL(h1_imag, x4);
#ifdef __ELPA_USE_FMA__
		q4 = _SSE_ADD(q4, _SSE_MADDSUB(h1_real, x4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#else
		q4 = _SSE_ADD(q4, _SSE_ADDSUB( _SSE_MUL(h1_real, x4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#endif
		tmp5 = _SSE_MUL(h1_imag, x5);
#ifdef __ELPA_USE_FMA__
		q5 = _SSE_ADD(q5, _SSE_MADDSUB(h1_real, x5, _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1))));
#else
		q5 = _SSE_ADD(q5, _SSE_ADDSUB( _SSE_MUL(h1_real, x5), _SSE_SHUFFLE(tmp5, tmp5, _MM_SHUFFLE2(0,1))));
#endif
		tmp6 = _SSE_MUL(h1_imag, x6);
#ifdef __ELPA_USE_FMA__
		q6 = _SSE_ADD(q6, _SSE_MADDSUB(h1_real, x6, _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1))));
#else
		q6 = _SSE_ADD(q6, _SSE_ADDSUB( _SSE_MUL(h1_real, x6), _SSE_SHUFFLE(tmp6, tmp6, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

		_SSE_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_SSE_STORE(&q_dbl[(2*i*ldq)+offset], q2);
		_SSE_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
#ifdef DOUBLE_PRECISION_COMPLEX
		_SSE_STORE(&q_dbl[(2*i*ldq)+3*offset], q4);
		_SSE_STORE(&q_dbl[(2*i*ldq)+4*offset], q5);
		_SSE_STORE(&q_dbl[(2*i*ldq)+5*offset], q6);
#endif
	}
}

#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_4_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_4_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq)
#endif
{
#ifdef DOUBLE_PRECISION_COMPLEX
	double* q_dbl = (double*)q;
	double* hh_dbl = (double*)hh;
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	float* q_dbl = (float*)q;
	float* hh_dbl = (float*)hh;
#endif
	__SSE_DATATYPE x1, x2, x3, x4;
	__SSE_DATATYPE q1, q2, q3, q4;
	__SSE_DATATYPE h1_real, h1_imag;
	__SSE_DATATYPE tmp1, tmp2, tmp3, tmp4;
	int i=0;
#ifdef DOUBLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi64x(0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000);
#endif

	x1 = _SSE_LOAD(&q_dbl[0]);
	x2 = _SSE_LOAD(&q_dbl[offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
	x3 = _SSE_LOAD(&q_dbl[2*offset]);
	x4 = _SSE_LOAD(&q_dbl[3*offset]);
#endif
	for (i = 1; i < nb; i++)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif
#ifndef __ELPA_USE_FMA__
		// conjugate
		h1_imag = _SSE_XOR(h1_imag, sign);
#endif

		q1 = _SSE_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _SSE_LOAD(&q_dbl[(2*i*ldq)+offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
		q3 = _SSE_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _SSE_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
#endif
		tmp1 = _SSE_MUL(h1_imag, q1);
#ifdef DOUBLE_PRECISION_COMPLEX

#ifdef __ELPA_USE_FMA__
		x1 = _SSE_ADD(x1, _mm_msubadd_pd(h1_real, q1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		x1 = _SSE_ADD(x1, _SSE_ADDSUB( _SSE_MUL(h1_real, q1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif

		tmp2 = _SSE_MUL(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
		x2 = _SSE_ADD(x2, _mm_msubadd_pd(h1_real, q2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		x2 = _SSE_ADD(x2, _SSE_ADDSUB( _SSE_MUL(h1_real, q2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		x1 = _mm_add_ps(x1, _mm_msubadd_ps(h1_real, q1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		x1 = _mm_add_ps(x1, _mm_addsub_ps( _mm_mul_ps(h1_real, q1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
		tmp2 = _mm_mul_ps(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
		x2 = _mm_add_ps(x2, _mm_msubadd_ps(h1_real, q2, _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#else
		x2 = _mm_add_ps(x2, _mm_addsub_ps( _mm_mul_ps(h1_real, q2), _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
		tmp3 = _SSE_MUL(h1_imag, q3);
#ifdef __ELPA_USE_FMA__
		x3 = _SSE_ADD(x3, _mm_msubadd_pd(h1_real, q3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#else
		x3 = _SSE_ADD(x3, _SSE_ADDSUB( _SSE_MUL(h1_real, q3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#endif
		tmp4 = _SSE_MUL(h1_imag, q4);
#ifdef __ELPA_USE_FMA__
		x4 = _SSE_ADD(x4, _mm_msubadd_pd(h1_real, q4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#else
		x4 = _SSE_ADD(x4, _SSE_ADDSUB( _SSE_MUL(h1_real, q4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */
	}

#ifdef DOUBLE_PRECISION_COMPLEX
	h1_real = _mm_loaddup_pd(&hh_dbl[0]);
	h1_imag = _mm_loaddup_pd(&hh_dbl[1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[0]) )));
	h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[1]) )));
#endif
	h1_real = _SSE_XOR(h1_real, sign);
	h1_imag = _SSE_XOR(h1_imag, sign);

	tmp1 = _SSE_MUL(h1_imag, x1);
#ifdef DOUBLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
	x1 = _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#else
	x1 = _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#endif
	tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
	x2 = _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#else
	x2 = _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX */
#ifdef __ELPA_USE_FMA__
	x1 = _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#else
	x1 = _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#endif
	tmp2 = _mm_mul_ps(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
	x2 = _mm_maddsub_ps(h1_real, x2, _mm_shuffle_ps(tmp2, tmp2, 0xb1));
#else
	x2 = _mm_addsub_ps( _mm_mul_ps(h1_real, x2), _mm_shuffle_ps(tmp2, tmp2, 0xb1));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
	tmp3 = _SSE_MUL(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
	x3 = _SSE_MADDSUB(h1_real, x3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1)));
#else
	x3 = _SSE_ADDSUB( _SSE_MUL(h1_real, x3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1)));
#endif
	tmp4 = _SSE_MUL(h1_imag, x4);
#ifdef __ELPA_USE_FMA__
	x4 = _SSE_MADDSUB(h1_real, x4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1)));
#else
	x4 = _SSE_ADDSUB( _SSE_MUL(h1_real, x4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

	q1 = _SSE_LOAD(&q_dbl[0]);
	q2 = _SSE_LOAD(&q_dbl[offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
	q3 = _SSE_LOAD(&q_dbl[2*offset]);
	q4 = _SSE_LOAD(&q_dbl[3*offset]);
#endif
	q1 = _SSE_ADD(q1, x1);
	q2 = _SSE_ADD(q2, x2);
#ifdef DOUBLE_PRECISION_COMPLEX
	q3 = _SSE_ADD(q3, x3);
	q4 = _SSE_ADD(q4, x4);
#endif
	_SSE_STORE(&q_dbl[0], q1);
	_SSE_STORE(&q_dbl[offset], q2);
#ifdef DOUBLE_PRECISION_COMPLEX
	_SSE_STORE(&q_dbl[2*offset], q3);
	_SSE_STORE(&q_dbl[3*offset], q4);
#endif
	for (i = 1; i < nb; i++)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif
		q1 = _SSE_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _SSE_LOAD(&q_dbl[(2*i*ldq)+offset]);
#ifdef DOUBLE_PRECISION_COMPLEX
		q3 = _SSE_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _SSE_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
#endif
		tmp1 = _SSE_MUL(h1_imag, x1);

#ifdef DOUBLE_PRECISION_COMPLEX

#ifdef __ELPA_USE_FMA__
		q1 = _SSE_ADD(q1, _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		q1 = _SSE_ADD(q1, _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif
		tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
		q2 = _SSE_ADD(q2, _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		q2 = _SSE_ADD(q2, _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX */
#ifdef __ELPA_USE_FMA__
		q1 = _mm_add_ps(q1, _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		q1 = _mm_add_ps(q1, _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
		tmp2 = _mm_mul_ps(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
		q2 = _mm_add_ps(q2, _mm_maddsub_ps(h1_real, x2, _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#else
		q2 = _mm_add_ps(q2, _mm_addsub_ps( _mm_mul_ps(h1_real, x2), _mm_shuffle_ps(tmp2, tmp2, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
		tmp3 = _SSE_MUL(h1_imag, x3);
#ifdef __ELPA_USE_FMA__
		q3 = _SSE_ADD(q3, _SSE_MADDSUB(h1_real, x3, _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#else
		q3 = _SSE_ADD(q3, _SSE_ADDSUB( _SSE_MUL(h1_real, x3), _SSE_SHUFFLE(tmp3, tmp3, _MM_SHUFFLE2(0,1))));
#endif
		tmp4 = _SSE_MUL(h1_imag, x4);
#ifdef __ELPA_USE_FMA__
		q4 = _SSE_ADD(q4, _SSE_MADDSUB(h1_real, x4, _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#else
		q4 = _SSE_ADD(q4, _SSE_ADDSUB( _SSE_MUL(h1_real, x4), _SSE_SHUFFLE(tmp4, tmp4, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

		_SSE_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_SSE_STORE(&q_dbl[(2*i*ldq)+offset], q2);
#ifdef DOUBLE_PRECISION_COMPLEX
		_SSE_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
		_SSE_STORE(&q_dbl[(2*i*ldq)+3*offset], q4);
#endif
	}
}

#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_2_SSE_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_2_SSE_1hv_single(float complex* q, float complex* hh, int nb, int ldq)
#endif
{

#ifdef DOUBLE_PRECISION_COMPLEX
	double* q_dbl = (double*)q;
	double* hh_dbl = (double*)hh;
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	float* q_dbl = (float*)q;
	float* hh_dbl = (float*)hh;
#endif
	__SSE_DATATYPE x1, x2;
	__SSE_DATATYPE q1, q2;
	__SSE_DATATYPE h1_real, h1_imag;
	__SSE_DATATYPE tmp1, tmp2;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi64x(0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__SSE_DATATYPE sign = (__SSE_DATATYPE)_mm_set_epi32(0x80000000, 0x80000000, 0x80000000, 0x80000000);
#endif
	x1 = _SSE_LOAD(&q_dbl[0]);
#ifdef DOUBLE_PRECISION_COMPLEX
	x2 = _SSE_LOAD(&q_dbl[offset]);
#endif
	for (i = 1; i < nb; i++)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif
#ifndef __ELPA_USE_FMA__
		// conjugate
		h1_imag = _SSE_XOR(h1_imag, sign);
#endif

		q1 = _SSE_LOAD(&q_dbl[(2*i*ldq)+0]);
#ifdef DOUBLE_PRECISION_COMPLEX
		q2 = _SSE_LOAD(&q_dbl[(2*i*ldq)+offset]);
#endif
		tmp1 = _SSE_MUL(h1_imag, q1);
#ifdef DOUBLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		x1 = _SSE_ADD(x1, _mm_msubadd_pd(h1_real, q1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		x1 = _SSE_ADD(x1, _SSE_ADDSUB( _SSE_MUL(h1_real, q1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		x1 = _mm_add_ps(x1, _mm_msubadd_ps(h1_real, q1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		x1 = _mm_add_ps(x1, _mm_addsub_ps( _mm_mul_ps(h1_real, q1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
		tmp2 = _SSE_MUL(h1_imag, q2);
#ifdef __ELPA_USE_FMA__
		x2 = _SSE_ADD(x2, _mm_msubadd_pd(h1_real, q2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		x2 = _SSE_ADD(x2, _SSE_ADDSUB( _SSE_MUL(h1_real, q2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */
	}

#ifdef DOUBLE_PRECISION_COMPLEX
	h1_real = _mm_loaddup_pd(&hh_dbl[0]);
	h1_imag = _mm_loaddup_pd(&hh_dbl[1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[0]) )));
	h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[1]) )));
#endif
	h1_real = _SSE_XOR(h1_real, sign);
	h1_imag = _SSE_XOR(h1_imag, sign);

	tmp1 = _SSE_MUL(h1_imag, x1);
#ifdef DOUBLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
	x1 = _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#else
	x1 = _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
	x1 = _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#else
	x1 = _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
	tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
	x2 = _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#else
	x2 = _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1)));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */
	q1 = _SSE_LOAD(&q_dbl[0]);
#ifdef DOUBLE_PRECISION_COMPLEX
	q2 = _SSE_LOAD(&q_dbl[offset]);
#endif
	q1 = _SSE_ADD(q1, x1);
#ifdef DOUBLE_PRECISION_COMPLEX
	q2 = _SSE_ADD(q2, x2);
#endif
	_SSE_STORE(&q_dbl[0], q1);
#ifdef DOUBLE_PRECISION_COMPLEX
	_SSE_STORE(&q_dbl[offset], q2);
#endif
	for (i = 1; i < nb; i++)
	{
#ifdef DOUBLE_PRECISION_COMPLEX
		h1_real = _mm_loaddup_pd(&hh_dbl[i*2]);
		h1_imag = _mm_loaddup_pd(&hh_dbl[(i*2)+1]);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
		h1_real = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[i*2]) )));
		h1_imag = _mm_moveldup_ps(_mm_castpd_ps(_mm_loaddup_pd( (double *)(&hh_dbl[(i*2)+1]) )));
#endif

		q1 = _SSE_LOAD(&q_dbl[(2*i*ldq)+0]);
#ifdef DOUBLE_PRECISION_COMPLEX
		q2 = _SSE_LOAD(&q_dbl[(2*i*ldq)+offset]);
#endif
		tmp1 = _SSE_MUL(h1_imag, x1);

#ifdef DOUBLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		q1 = _SSE_ADD(q1, _SSE_MADDSUB(h1_real, x1, _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#else
		q1 = _SSE_ADD(q1, _SSE_ADDSUB( _SSE_MUL(h1_real, x1), _SSE_SHUFFLE(tmp1, tmp1, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#ifdef __ELPA_USE_FMA__
		q1 = _mm_add_ps(q1, _mm_maddsub_ps(h1_real, x1, _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#else
		q1 = _mm_add_ps(q1, _mm_addsub_ps( _mm_mul_ps(h1_real, x1), _mm_shuffle_ps(tmp1, tmp1, 0xb1)));
#endif
#endif /* SINGLE_PRECISION_COMPLEX */

#ifdef DOUBLE_PRECISION_COMPLEX
		tmp2 = _SSE_MUL(h1_imag, x2);
#ifdef __ELPA_USE_FMA__
		q2 = _SSE_ADD(q2, _SSE_MADDSUB(h1_real, x2, _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#else
		q2 = _SSE_ADD(q2, _SSE_ADDSUB( _SSE_MUL(h1_real, x2), _SSE_SHUFFLE(tmp2, tmp2, _MM_SHUFFLE2(0,1))));
#endif
#endif /* DOUBLE_PRECISION_COMPLEX */
		_SSE_STORE(&q_dbl[(2*i*ldq)+0], q1);
#ifdef DOUBLE_PRECISION_COMPLEX
		_SSE_STORE(&q_dbl[(2*i*ldq)+offset], q2);
#endif
	}
}