complex_avx512_1hv_template.c 33 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
//    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 (andreas.marek@mpcdf.mpg.de)
// --------------------------------------------------------------------------------------------------


#include "config-f90.h"

#include <complex.h>
#include <x86intrin.h>
53
#include <stdio.h>
Andreas Marek's avatar
Andreas Marek committed
54
#include <stdlib.h>
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

#define __forceinline __attribute__((always_inline))

#ifdef DOUBLE_PRECISION_COMPLEX
#define __AVX512_DATATYPE __m512d
#define _AVX512_LOAD _mm512_load_pd
#define _AVX512_STORE _mm512_store_pd
#define _AVX512_SET1 _mm512_set1_pd
#define _AVX512_MUL _mm512_mul_pd
#define _AVX512_ADD _mm512_add_pd
#define _AVX512_SHUFFLE _mm512_shuffle_pd
#define _AVX512_XOR _mm512_xor_pd
#define _AVX512_XOR_EPI _mm512_xor_epi64
#define _SHUFFLE 0x55

#ifdef HAVE_AVX512

#define __ELPA_USE_FMA__
#define _mm512_FMADDSUB_pd(a,b,c) _mm512_fmaddsub_pd(a,b,c)
#define _mm512_FMSUBADD_pd(a,b,c) _mm512_fmsubadd_pd(a,b,c)

#endif

#define _AVX512_FMADDSUB _mm512_FMADDSUB_pd
#define _AVX512_FMSUBADD _mm512_FMSUBADD_pd
#endif /* DOUBLE_PRECISION_COMPLEX */

#ifdef SINGLE_PRECISION_COMPLEX
#define __AVX512_DATATYPE __m512
#define _AVX512_LOAD _mm512_load_ps
#define _AVX512_STORE _mm512_store_ps
#define _AVX512_SET1 _mm512_set1_ps
#define _AVX512_MUL _mm512_mul_ps
#define _AVX512_ADD _mm512_add_ps
#define _AVX512_SHUFFLE _mm512_shuffle_ps
#define _AVX512_XOR _mm512_xor_ps
#define _AVX512_XOR_EPI _mm512_xor_epi32
#define _SHUFFLE 0xb1

#ifdef HAVE_AVX512

#define __ELPA_USE_FMA__
#define _mm512_FMADDSUB_ps(a,b,c) _mm512_fmaddsub_ps(a,b,c)
#define _mm512_FMSUBADD_ps(a,b,c) _mm512_fmsubadd_ps(a,b,c)

#endif

#define _AVX512_FMADDSUB _mm512_FMADDSUB_ps
#define _AVX512_FMSUBADD _mm512_FMSUBADD_ps
#endif /* SINGLE_PRECISION_COMPLEX */


//Forward declaration
#ifdef DOUBLE_PRECISION_COMPLEX
static  __forceinline void hh_trafo_complex_kernel_24_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
110
static  __forceinline void hh_trafo_complex_kernel_20_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
111
static  __forceinline void hh_trafo_complex_kernel_16_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
112
static  __forceinline void hh_trafo_complex_kernel_12_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
113
static  __forceinline void hh_trafo_complex_kernel_8_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
114
static  __forceinline void hh_trafo_complex_kernel_4_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq);
115 116 117 118
#endif

#ifdef SINGLE_PRECISION_COMPLEX
static  __forceinline void hh_trafo_complex_kernel_48_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
119
static  __forceinline void hh_trafo_complex_kernel_40_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
120
static  __forceinline void hh_trafo_complex_kernel_32_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
121
static  __forceinline void hh_trafo_complex_kernel_24_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
122
static  __forceinline void hh_trafo_complex_kernel_16_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
123
static  __forceinline void hh_trafo_complex_kernel_8_AVX512_1hv_single(float complex* q, float complex* hh, int nb, int ldq);
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
#endif


/*
!f>#if defined(HAVE_AVX512)
!f> interface
!f>   subroutine single_hh_trafo_complex_avx512_1hv_double(q, hh, pnb, pnq, pldq) &
!f>                             bind(C, name="single_hh_trafo_complex_avx512_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>#if defined(HAVE_AVX512)
!f> interface
!f>   subroutine single_hh_trafo_complex_avx512_1hv_single(q, hh, pnb, pnq, pldq) &
!f>                             bind(C, name="single_hh_trafo_complex_avx512_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
*/

#ifdef DOUBLE_PRECISION_COMPLEX
void single_hh_trafo_complex_avx512_1hv_double(double complex* q, double complex* hh, int* pnb, int* pnq, int* pldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
void single_hh_trafo_complex_avx512_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;
167
	int worked_on;
168 169
	//int ldh = *pldh;

170 171
	worked_on = 0;

172
#ifdef DOUBLE_PRECISION_COMPLEX
173
	for (i = 0; i < nq-20; i+=24)
174 175
	{
		hh_trafo_complex_kernel_24_AVX512_1hv_double(&q[i], hh, nb, ldq);
176
		worked_on += 24;
177 178
	}
#endif
179

180
#ifdef SINGLE_PRECISION_COMPLEX
181
	for (i = 0; i < nq-40; i+=48)
182 183
	{
		hh_trafo_complex_kernel_48_AVX512_1hv_single(&q[i], hh, nb, ldq);
184
		worked_on += 48;
185 186 187 188 189 190
	}
#endif
	if (nq == i)
	{
		return;
	}
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207

#ifdef DOUBLE_PRECISION_COMPLEX
	if (nq-i == 20)
	{
		hh_trafo_complex_kernel_20_AVX512_1hv_double(&q[i], hh, nb, ldq);
		worked_on += 20;
	}
#endif

#ifdef SINGLE_PRECISION_COMPLEX
	if (nq-i == 40)
	{
		hh_trafo_complex_kernel_40_AVX512_1hv_single(&q[i], hh, nb, ldq);
		worked_on += 40;
	}
#endif

208 209 210 211
#ifdef DOUBLE_PRECISION_COMPLEX
	if (nq-i == 16)
	{
		hh_trafo_complex_kernel_16_AVX512_1hv_double(&q[i], hh, nb, ldq);
212
		worked_on += 16;
213 214
	}
#endif
215

216 217 218 219
#ifdef SINGLE_PRECISION_COMPLEX
	if (nq-i == 32)
	{
		hh_trafo_complex_kernel_32_AVX512_1hv_single(&q[i], hh, nb, ldq);
220
		worked_on += 32;
221 222
	}
#endif
223 224 225

#ifdef DOUBLE_PRECISION_COMPLEX
	if (nq-i == 12)
226
	{
227 228 229 230 231 232 233 234 235 236 237 238 239
		hh_trafo_complex_kernel_12_AVX512_1hv_double(&q[i], hh, nb, ldq);
		worked_on += 12;
	}
#endif

#ifdef SINGLE_PRECISION_COMPLEX
	if (nq-i == 24)
	{
		hh_trafo_complex_kernel_24_AVX512_1hv_single(&q[i], hh, nb, ldq);
		worked_on += 24;
	}
#endif

240
#ifdef DOUBLE_PRECISION_COMPLEX
241 242
	if (nq-i == 8)
	{
243
		hh_trafo_complex_kernel_8_AVX512_1hv_double(&q[i], hh, nb, ldq);
244 245
		worked_on += 8;
	}
246
#endif
247

248
#ifdef SINGLE_PRECISION_COMPLEX
249 250
	if (nq-i == 16)
	{
251
		hh_trafo_complex_kernel_16_AVX512_1hv_single(&q[i], hh, nb, ldq);
252 253
		worked_on += 16;
	}
254
#endif
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272

#ifdef DOUBLE_PRECISION_COMPLEX
	if (nq-i == 4)
	{
		hh_trafo_complex_kernel_4_AVX512_1hv_double(&q[i], hh, nb, ldq);
		worked_on += 4;
	}
#endif

#ifdef SINGLE_PRECISION_COMPLEX
	if (nq-i == 8)
	{
		hh_trafo_complex_kernel_8_AVX512_1hv_single(&q[i], hh, nb, ldq);
		worked_on += 8;
	}
#endif
	if (worked_on != nq)
	{
273 274
	     printf("Error in complex AVX512 BLOCK 1 kernel \n");
	     abort();
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
	}
}

#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_24_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_48_AVX512_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
	__AVX512_DATATYPE x1, x2, x3, x4, x5, x6;
	__AVX512_DATATYPE q1, q2, q3, q4, q5, q6;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
#define offset 8
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define offset 16
#endif


	x1 = _AVX512_LOAD(&q_dbl[0]);    // complex 1, 2, 3, 4
	x2 = _AVX512_LOAD(&q_dbl[offset]);    // complex 5, 6, 7, 8
	x3 = _AVX512_LOAD(&q_dbl[2*offset]);   // complex 9, 10, 11, 12
	x4 = _AVX512_LOAD(&q_dbl[3*offset]);   // complex 13, 14, 15, 16
	x5 = _AVX512_LOAD(&q_dbl[4*offset]);   // complex 17, 18, 19, 20
	x6 = _AVX512_LOAD(&q_dbl[5*offset]);   // complex 21, 22, 23, 24

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
		q5 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+4*offset]);
		q6 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+5*offset]);

		tmp1 = _AVX512_MUL(h1_imag, q1);

		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, q2);

		x2 = _AVX512_ADD(x2, _AVX512_FMSUBADD(h1_real, q2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, q3);

		x3 = _AVX512_ADD(x3, _AVX512_FMSUBADD(h1_real, q3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, q4);

		x4 = _AVX512_ADD(x4, _AVX512_FMSUBADD(h1_real, q4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));

		tmp5 = _AVX512_MUL(h1_imag, q5);

		x5 = _AVX512_ADD(x5, _AVX512_FMSUBADD(h1_real, q5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE)));

		tmp6 = _AVX512_MUL(h1_imag, q6);

		x6 = _AVX512_ADD(x6, _AVX512_FMSUBADD(h1_real, q6, _AVX512_SHUFFLE(tmp6, tmp6, _SHUFFLE)));
	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);

	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	tmp2 = _AVX512_MUL(h1_imag, x2);

	x2 = _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE));

	tmp3 = _AVX512_MUL(h1_imag, x3);

	x3 = _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE));

	tmp4 = _AVX512_MUL(h1_imag, x4);

	x4 = _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE));

	tmp5 = _AVX512_MUL(h1_imag, x5);

	x5 = _AVX512_FMADDSUB(h1_real, x5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE));

	tmp6 = _AVX512_MUL(h1_imag, x6);

	x6 = _AVX512_FMADDSUB(h1_real, x6, _AVX512_SHUFFLE(tmp6, tmp6, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);
	q2 = _AVX512_LOAD(&q_dbl[offset]);
	q3 = _AVX512_LOAD(&q_dbl[2*offset]);
	q4 = _AVX512_LOAD(&q_dbl[3*offset]);
	q5 = _AVX512_LOAD(&q_dbl[4*offset]);
	q6 = _AVX512_LOAD(&q_dbl[5*offset]);

	q1 = _AVX512_ADD(q1, x1);
	q2 = _AVX512_ADD(q2, x2);
	q3 = _AVX512_ADD(q3, x3);
	q4 = _AVX512_ADD(q4, x4);
	q5 = _AVX512_ADD(q5, x5);
	q6 = _AVX512_ADD(q6, x6);

	_AVX512_STORE(&q_dbl[0], q1);
	_AVX512_STORE(&q_dbl[offset], q2);
	_AVX512_STORE(&q_dbl[2*offset], q3);
	_AVX512_STORE(&q_dbl[3*offset], q4);
	_AVX512_STORE(&q_dbl[4*offset], q5);
	_AVX512_STORE(&q_dbl[5*offset], q6);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
		q5 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+4*offset]);
		q6 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+5*offset]);

		tmp1 = _AVX512_MUL(h1_imag, x1);

		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, x2);

		q2 = _AVX512_ADD(q2, _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, x3);

		q3 = _AVX512_ADD(q3, _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, x4);

		q4 = _AVX512_ADD(q4, _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));

		tmp5 = _AVX512_MUL(h1_imag, x5);

		q5 = _AVX512_ADD(q5, _AVX512_FMADDSUB(h1_real, x5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE)));

		tmp6 = _AVX512_MUL(h1_imag, x6);

		q6 = _AVX512_ADD(q6, _AVX512_FMADDSUB(h1_real, x6, _AVX512_SHUFFLE(tmp6, tmp6, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+offset], q2);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+3*offset], q4);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+4*offset], q5);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+5*offset], q6);
	}
}

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
#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_20_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_40_AVX512_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
	__AVX512_DATATYPE x1, x2, x3, x4, x5, x6;
	__AVX512_DATATYPE q1, q2, q3, q4, q5, q6;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2, tmp3, tmp4, tmp5, tmp6;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

#ifdef DOUBLE_PRECISION_COMPLEX
#define offset 8
#endif
#ifdef SINGLE_PRECISION_COMPLEX
#define offset 16
#endif


	x1 = _AVX512_LOAD(&q_dbl[0]);    // complex 1, 2, 3, 4
	x2 = _AVX512_LOAD(&q_dbl[offset]);    // complex 5, 6, 7, 8
	x3 = _AVX512_LOAD(&q_dbl[2*offset]);   // complex 9, 10, 11, 12
	x4 = _AVX512_LOAD(&q_dbl[3*offset]);   // complex 13, 14, 15, 16
	x5 = _AVX512_LOAD(&q_dbl[4*offset]);   // complex 17, 18, 19, 20

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
		q5 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+4*offset]);

		tmp1 = _AVX512_MUL(h1_imag, q1);

		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, q2);

		x2 = _AVX512_ADD(x2, _AVX512_FMSUBADD(h1_real, q2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, q3);

		x3 = _AVX512_ADD(x3, _AVX512_FMSUBADD(h1_real, q3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, q4);

		x4 = _AVX512_ADD(x4, _AVX512_FMSUBADD(h1_real, q4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));

		tmp5 = _AVX512_MUL(h1_imag, q5);

		x5 = _AVX512_ADD(x5, _AVX512_FMSUBADD(h1_real, q5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE)));

	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);

	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	tmp2 = _AVX512_MUL(h1_imag, x2);

	x2 = _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE));

	tmp3 = _AVX512_MUL(h1_imag, x3);

	x3 = _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE));

	tmp4 = _AVX512_MUL(h1_imag, x4);

	x4 = _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE));

	tmp5 = _AVX512_MUL(h1_imag, x5);

	x5 = _AVX512_FMADDSUB(h1_real, x5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);
	q2 = _AVX512_LOAD(&q_dbl[offset]);
	q3 = _AVX512_LOAD(&q_dbl[2*offset]);
	q4 = _AVX512_LOAD(&q_dbl[3*offset]);
	q5 = _AVX512_LOAD(&q_dbl[4*offset]);

	q1 = _AVX512_ADD(q1, x1);
	q2 = _AVX512_ADD(q2, x2);
	q3 = _AVX512_ADD(q3, x3);
	q4 = _AVX512_ADD(q4, x4);
	q5 = _AVX512_ADD(q5, x5);

	_AVX512_STORE(&q_dbl[0], q1);
	_AVX512_STORE(&q_dbl[offset], q2);
	_AVX512_STORE(&q_dbl[2*offset], q3);
	_AVX512_STORE(&q_dbl[3*offset], q4);
	_AVX512_STORE(&q_dbl[4*offset], q5);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);
		q5 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+4*offset]);

		tmp1 = _AVX512_MUL(h1_imag, x1);

		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, x2);

		q2 = _AVX512_ADD(q2, _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, x3);

		q3 = _AVX512_ADD(q3, _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, x4);

		q4 = _AVX512_ADD(q4, _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));

		tmp5 = _AVX512_MUL(h1_imag, x5);

		q5 = _AVX512_ADD(q5, _AVX512_FMADDSUB(h1_real, x5, _AVX512_SHUFFLE(tmp5, tmp5, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+offset], q2);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+3*offset], q4);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+4*offset], q5);
	}
}


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
#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_16_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_32_AVX512_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

	__AVX512_DATATYPE x1, x2, x3, x4;
	__AVX512_DATATYPE q1, q2, q3, q4;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2, tmp3, tmp4;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

	x1 = _AVX512_LOAD(&q_dbl[0]);   // complex 1 2 3 4
	x2 = _AVX512_LOAD(&q_dbl[offset]);
	x3 = _AVX512_LOAD(&q_dbl[2*offset]);
	x4 = _AVX512_LOAD(&q_dbl[3*offset]);  // comlex 13 14 15 16

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);

		tmp1 = _AVX512_MUL(h1_imag, q1);

		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, q2);

		x2 = _AVX512_ADD(x2, _AVX512_FMSUBADD(h1_real, q2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, q3);

		x3 = _AVX512_ADD(x3, _AVX512_FMSUBADD(h1_real, q3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, q4);

		x4 = _AVX512_ADD(x4, _AVX512_FMSUBADD(h1_real, q4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));
	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);

	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	tmp2 = _AVX512_MUL(h1_imag, x2);

	x2 = _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE));

	tmp3 = _AVX512_MUL(h1_imag, x3);

	x3 = _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE));

	tmp4 = _AVX512_MUL(h1_imag, x4);

	x4 = _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);
	q2 = _AVX512_LOAD(&q_dbl[offset]);
	q3 = _AVX512_LOAD(&q_dbl[2*offset]);
	q4 = _AVX512_LOAD(&q_dbl[3*offset]);

	q1 = _AVX512_ADD(q1, x1);
	q2 = _AVX512_ADD(q2, x2);
	q3 = _AVX512_ADD(q3, x3);
	q4 = _AVX512_ADD(q4, x4);

	_AVX512_STORE(&q_dbl[0], q1);
	_AVX512_STORE(&q_dbl[offset], q2);
	_AVX512_STORE(&q_dbl[2*offset], q3);
	_AVX512_STORE(&q_dbl[3*offset], q4);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);
		q4 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+3*offset]);

		tmp1 = _AVX512_MUL(h1_imag, x1);

		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, x2);

		q2 = _AVX512_ADD(q2, _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, x3);

		q3 = _AVX512_ADD(q3, _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		tmp4 = _AVX512_MUL(h1_imag, x4);

		q4 = _AVX512_ADD(q4, _AVX512_FMADDSUB(h1_real, x4, _AVX512_SHUFFLE(tmp4, tmp4, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+offset], q2);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+3*offset], q4);
	}
}

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
#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_12_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_24_AVX512_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

	__AVX512_DATATYPE x1, x2, x3, x4;
	__AVX512_DATATYPE q1, q2, q3, q4;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2, tmp3, tmp4;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

	x1 = _AVX512_LOAD(&q_dbl[0]);   // complex 1 2 3 4
	x2 = _AVX512_LOAD(&q_dbl[offset]);
	x3 = _AVX512_LOAD(&q_dbl[2*offset]);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);

		tmp1 = _AVX512_MUL(h1_imag, q1);

		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, q2);

		x2 = _AVX512_ADD(x2, _AVX512_FMSUBADD(h1_real, q2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, q3);

		x3 = _AVX512_ADD(x3, _AVX512_FMSUBADD(h1_real, q3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
        h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
        h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);

	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	tmp2 = _AVX512_MUL(h1_imag, x2);

	x2 = _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE));

	tmp3 = _AVX512_MUL(h1_imag, x3);

	x3 = _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);
	q2 = _AVX512_LOAD(&q_dbl[offset]);
	q3 = _AVX512_LOAD(&q_dbl[2*offset]);

	q1 = _AVX512_ADD(q1, x1);
	q2 = _AVX512_ADD(q2, x2);
	q3 = _AVX512_ADD(q3, x3);

	_AVX512_STORE(&q_dbl[0], q1);
	_AVX512_STORE(&q_dbl[offset], q2);
	_AVX512_STORE(&q_dbl[2*offset], q3);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);
		q3 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+2*offset]);

		tmp1 = _AVX512_MUL(h1_imag, x1);

		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, x2);

		q2 = _AVX512_ADD(q2, _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		tmp3 = _AVX512_MUL(h1_imag, x3);

		q3 = _AVX512_ADD(q3, _AVX512_FMADDSUB(h1_real, x3, _AVX512_SHUFFLE(tmp3, tmp3, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+offset], q2);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+2*offset], q3);
	}
}


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
#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_8_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_16_AVX512_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
	__AVX512_DATATYPE x1, x2;
	__AVX512_DATATYPE q1, q2;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

	x1 = _AVX512_LOAD(&q_dbl[0]);
	x2 = _AVX512_LOAD(&q_dbl[offset]);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);

		tmp1 = _AVX512_MUL(h1_imag, q1);
		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
		tmp2 = _AVX512_MUL(h1_imag, q2);
		x2 = _AVX512_ADD(x2, _AVX512_FMSUBADD(h1_real, q2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));
	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
	h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
	h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
	h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);
	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	tmp2 = _AVX512_MUL(h1_imag, x2);

	x2 = _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);
	q2 = _AVX512_LOAD(&q_dbl[offset]);

	q1 = _AVX512_ADD(q1, x1);
	q2 = _AVX512_ADD(q2, x2);

	_AVX512_STORE(&q_dbl[0], q1);
	_AVX512_STORE(&q_dbl[offset], q2);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);
		q2 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+offset]);

		tmp1 = _AVX512_MUL(h1_imag, x1);
		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		tmp2 = _AVX512_MUL(h1_imag, x2);

		q2 = _AVX512_ADD(q2, _AVX512_FMADDSUB(h1_real, x2, _AVX512_SHUFFLE(tmp2, tmp2, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
		_AVX512_STORE(&q_dbl[(2*i*ldq)+offset], q2);
	}
}
979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057


#ifdef DOUBLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_4_AVX512_1hv_double(double complex* q, double complex* hh, int nb, int ldq)
#endif
#ifdef SINGLE_PRECISION_COMPLEX
static __forceinline void hh_trafo_complex_kernel_8_AVX512_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
	__AVX512_DATATYPE x1, x2;
	__AVX512_DATATYPE q1, q2;
	__AVX512_DATATYPE h1_real, h1_imag;
	__AVX512_DATATYPE tmp1, tmp2;
	int i=0;

#ifdef DOUBLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set_epi64(0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	__AVX512_DATATYPE sign = (__AVX512_DATATYPE)_mm512_set1_epi32(0x80000000);
#endif

	x1 = _AVX512_LOAD(&q_dbl[0]);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);

		tmp1 = _AVX512_MUL(h1_imag, q1);
		x1 = _AVX512_ADD(x1, _AVX512_FMSUBADD(h1_real, q1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));
	}

	h1_real = _AVX512_SET1(hh_dbl[0]);
	h1_imag = _AVX512_SET1(hh_dbl[1]);

#ifdef DOUBLE_PRECISION_COMPLEX
	h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
	h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif
#ifdef SINGLE_PRECISION_COMPLEX
	h1_real = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_real, (__m512i) sign);
	h1_imag = (__AVX512_DATATYPE) _AVX512_XOR_EPI((__m512i) h1_imag, (__m512i) sign);
#endif

	tmp1 = _AVX512_MUL(h1_imag, x1);
	x1 = _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE));

	q1 = _AVX512_LOAD(&q_dbl[0]);

	q1 = _AVX512_ADD(q1, x1);

	_AVX512_STORE(&q_dbl[0], q1);

	for (i = 1; i < nb; i++)
	{
		h1_real = _AVX512_SET1(hh_dbl[i*2]);
		h1_imag = _AVX512_SET1(hh_dbl[(i*2)+1]);

		q1 = _AVX512_LOAD(&q_dbl[(2*i*ldq)+0]);

		tmp1 = _AVX512_MUL(h1_imag, x1);
		q1 = _AVX512_ADD(q1, _AVX512_FMADDSUB(h1_real, x1, _AVX512_SHUFFLE(tmp1, tmp1, _SHUFFLE)));

		_AVX512_STORE(&q_dbl[(2*i*ldq)+0], q1);
	}
}