2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2014,2015,2016,2017,2018,2019, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 #ifndef GMX_SIMD_IMPL_X86_AVX_512_SIMD_DOUBLE_H
37 #define GMX_SIMD_IMPL_X86_AVX_512_SIMD_DOUBLE_H
44 #include <immintrin.h>
46 #include "gromacs/math/utilities.h"
47 #include "gromacs/utility/basedefinitions.h"
49 #include "impl_x86_avx_512_general.h"
50 #include "impl_x86_avx_512_simd_float.h"
60 SimdDouble(double d) : simdInternal_(_mm512_set1_pd(d)) {}
62 // Internal utility constructor to simplify return statements
63 SimdDouble(__m512d simd) : simdInternal_(simd) {}
65 __m512d simdInternal_;
73 SimdDInt32(std::int32_t i) : simdInternal_(_mm256_set1_epi32(i)) {}
75 // Internal utility constructor to simplify return statements
76 SimdDInt32(__m256i simd) : simdInternal_(simd) {}
78 __m256i simdInternal_;
86 // Internal utility constructor to simplify return statements
87 SimdDBool(__mmask8 simd) : simdInternal_(simd) {}
89 __mmask8 simdInternal_;
97 // Internal utility constructor to simplify return statements
98 SimdDIBool(__mmask16 simd) : simdInternal_(simd) {}
100 __mmask16 simdInternal_;
103 static inline SimdDouble gmx_simdcall simdLoad(const double* m, SimdDoubleTag = {})
105 assert(std::size_t(m) % 64 == 0);
106 return { _mm512_load_pd(m) };
109 static inline void gmx_simdcall store(double* m, SimdDouble a)
111 assert(std::size_t(m) % 64 == 0);
112 _mm512_store_pd(m, a.simdInternal_);
115 static inline SimdDouble gmx_simdcall simdLoadU(const double* m, SimdDoubleTag = {})
117 return { _mm512_loadu_pd(m) };
120 static inline void gmx_simdcall storeU(double* m, SimdDouble a)
122 _mm512_storeu_pd(m, a.simdInternal_);
125 static inline SimdDouble gmx_simdcall setZeroD()
127 return { _mm512_setzero_pd() };
130 static inline SimdDInt32 gmx_simdcall simdLoad(const std::int32_t* m, SimdDInt32Tag)
132 assert(std::size_t(m) % 32 == 0);
133 return { _mm256_load_si256(reinterpret_cast<const __m256i*>(m)) };
136 static inline void gmx_simdcall store(std::int32_t* m, SimdDInt32 a)
138 assert(std::size_t(m) % 32 == 0);
139 _mm256_store_si256(reinterpret_cast<__m256i*>(m), a.simdInternal_);
142 static inline SimdDInt32 gmx_simdcall simdLoadU(const std::int32_t* m, SimdDInt32Tag)
144 return { _mm256_loadu_si256(reinterpret_cast<const __m256i*>(m)) };
147 static inline void gmx_simdcall storeU(std::int32_t* m, SimdDInt32 a)
149 _mm256_storeu_si256(reinterpret_cast<__m256i*>(m), a.simdInternal_);
152 static inline SimdDInt32 gmx_simdcall setZeroDI()
154 return { _mm256_setzero_si256() };
157 static inline SimdDouble gmx_simdcall operator&(SimdDouble a, SimdDouble b)
159 return { _mm512_castsi512_pd(_mm512_and_epi32(_mm512_castpd_si512(a.simdInternal_),
160 _mm512_castpd_si512(b.simdInternal_))) };
163 static inline SimdDouble gmx_simdcall andNot(SimdDouble a, SimdDouble b)
165 return { _mm512_castsi512_pd(_mm512_andnot_epi32(_mm512_castpd_si512(a.simdInternal_),
166 _mm512_castpd_si512(b.simdInternal_))) };
169 static inline SimdDouble gmx_simdcall operator|(SimdDouble a, SimdDouble b)
171 return { _mm512_castsi512_pd(_mm512_or_epi32(_mm512_castpd_si512(a.simdInternal_),
172 _mm512_castpd_si512(b.simdInternal_))) };
175 static inline SimdDouble gmx_simdcall operator^(SimdDouble a, SimdDouble b)
177 return { _mm512_castsi512_pd(_mm512_xor_epi32(_mm512_castpd_si512(a.simdInternal_),
178 _mm512_castpd_si512(b.simdInternal_))) };
181 static inline SimdDouble gmx_simdcall operator+(SimdDouble a, SimdDouble b)
183 return { _mm512_add_pd(a.simdInternal_, b.simdInternal_) };
186 static inline SimdDouble gmx_simdcall operator-(SimdDouble a, SimdDouble b)
188 return { _mm512_sub_pd(a.simdInternal_, b.simdInternal_) };
191 static inline SimdDouble gmx_simdcall operator-(SimdDouble x)
193 return { _mm512_castsi512_pd(_mm512_xor_epi32(_mm512_castpd_si512(x.simdInternal_),
194 _mm512_castpd_si512(_mm512_set1_pd(GMX_DOUBLE_NEGZERO)))) };
197 static inline SimdDouble gmx_simdcall operator*(SimdDouble a, SimdDouble b)
199 return { _mm512_mul_pd(a.simdInternal_, b.simdInternal_) };
202 static inline SimdDouble gmx_simdcall fma(SimdDouble a, SimdDouble b, SimdDouble c)
204 return { _mm512_fmadd_pd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
207 static inline SimdDouble gmx_simdcall fms(SimdDouble a, SimdDouble b, SimdDouble c)
209 return { _mm512_fmsub_pd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
212 static inline SimdDouble gmx_simdcall fnma(SimdDouble a, SimdDouble b, SimdDouble c)
214 return { _mm512_fnmadd_pd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
217 static inline SimdDouble gmx_simdcall fnms(SimdDouble a, SimdDouble b, SimdDouble c)
219 return { _mm512_fnmsub_pd(a.simdInternal_, b.simdInternal_, c.simdInternal_) };
222 // Override for AVX-512-KNL
223 #if GMX_SIMD_X86_AVX_512
224 static inline SimdDouble gmx_simdcall rsqrt(SimdDouble x)
226 return { _mm512_rsqrt14_pd(x.simdInternal_) };
229 static inline SimdDouble gmx_simdcall rcp(SimdDouble x)
231 return { _mm512_rcp14_pd(x.simdInternal_) };
235 static inline SimdDouble gmx_simdcall maskAdd(SimdDouble a, SimdDouble b, SimdDBool m)
237 return { _mm512_mask_add_pd(a.simdInternal_, m.simdInternal_, a.simdInternal_, b.simdInternal_) };
240 static inline SimdDouble gmx_simdcall maskzMul(SimdDouble a, SimdDouble b, SimdDBool m)
242 return { _mm512_maskz_mul_pd(m.simdInternal_, a.simdInternal_, b.simdInternal_) };
245 static inline SimdDouble gmx_simdcall maskzFma(SimdDouble a, SimdDouble b, SimdDouble c, SimdDBool m)
247 return { _mm512_maskz_fmadd_pd(m.simdInternal_, a.simdInternal_, b.simdInternal_, c.simdInternal_) };
250 // Override for AVX-512-KNL
251 #if GMX_SIMD_X86_AVX_512
252 static inline SimdDouble gmx_simdcall maskzRsqrt(SimdDouble x, SimdDBool m)
254 return { _mm512_maskz_rsqrt14_pd(m.simdInternal_, x.simdInternal_) };
257 static inline SimdDouble gmx_simdcall maskzRcp(SimdDouble x, SimdDBool m)
259 return { _mm512_maskz_rcp14_pd(m.simdInternal_, x.simdInternal_) };
263 static inline SimdDouble gmx_simdcall abs(SimdDouble x)
265 return { _mm512_castsi512_pd(_mm512_andnot_epi32(_mm512_castpd_si512(_mm512_set1_pd(GMX_DOUBLE_NEGZERO)),
266 _mm512_castpd_si512(x.simdInternal_))) };
269 static inline SimdDouble gmx_simdcall max(SimdDouble a, SimdDouble b)
271 return { _mm512_max_pd(a.simdInternal_, b.simdInternal_) };
274 static inline SimdDouble gmx_simdcall min(SimdDouble a, SimdDouble b)
276 return { _mm512_min_pd(a.simdInternal_, b.simdInternal_) };
279 static inline SimdDouble gmx_simdcall round(SimdDouble x)
281 return { _mm512_roundscale_pd(x.simdInternal_, 0) };
284 static inline SimdDouble gmx_simdcall trunc(SimdDouble x)
286 #if defined(__INTEL_COMPILER) || defined(__ECC)
287 return { _mm512_trunc_pd(x.simdInternal_) };
289 return { _mm512_cvtepi32_pd(_mm512_cvttpd_epi32(x.simdInternal_)) };
293 static inline SimdDouble frexp(SimdDouble value, SimdDInt32* exponent)
295 __m512d rExponent = _mm512_getexp_pd(value.simdInternal_);
296 __m256i iExponent = _mm512_cvtpd_epi32(rExponent);
298 exponent->simdInternal_ = _mm256_add_epi32(iExponent, _mm256_set1_epi32(1));
300 return { _mm512_getmant_pd(value.simdInternal_, _MM_MANT_NORM_p5_1, _MM_MANT_SIGN_src) };
303 template<MathOptimization opt = MathOptimization::Safe>
304 static inline SimdDouble ldexp(SimdDouble value, SimdDInt32 exponent)
306 const __m256i exponentBias = _mm256_set1_epi32(1023);
307 __m256i iExponent = _mm256_add_epi32(exponent.simdInternal_, exponentBias);
308 __m512i iExponent512;
310 if (opt == MathOptimization::Safe)
312 // Make sure biased argument is not negative
313 iExponent = _mm256_max_epi32(iExponent, _mm256_setzero_si256());
317 _mm512_permutexvar_epi32(_mm512_set_epi32(7, 7, 6, 6, 5, 5, 4, 4, 3, 3, 2, 2, 1, 1, 0, 0),
318 _mm512_castsi256_si512(iExponent));
320 _mm512_mask_slli_epi32(_mm512_setzero_epi32(), avx512Int2Mask(0xAAAA), iExponent512, 20);
321 return _mm512_mul_pd(_mm512_castsi512_pd(iExponent512), value.simdInternal_);
324 static inline double gmx_simdcall reduce(SimdDouble a)
326 __m512d x = a.simdInternal_;
327 x = _mm512_add_pd(x, _mm512_shuffle_f64x2(x, x, 0xEE));
328 x = _mm512_add_pd(x, _mm512_shuffle_f64x2(x, x, 0x11));
329 x = _mm512_add_pd(x, _mm512_permute_pd(x, 0x01));
330 return *reinterpret_cast<double*>(&x);
333 static inline SimdDBool gmx_simdcall operator==(SimdDouble a, SimdDouble b)
335 return { _mm512_cmp_pd_mask(a.simdInternal_, b.simdInternal_, _CMP_EQ_OQ) };
338 static inline SimdDBool gmx_simdcall operator!=(SimdDouble a, SimdDouble b)
340 return { _mm512_cmp_pd_mask(a.simdInternal_, b.simdInternal_, _CMP_NEQ_OQ) };
343 static inline SimdDBool gmx_simdcall operator<(SimdDouble a, SimdDouble b)
345 return { _mm512_cmp_pd_mask(a.simdInternal_, b.simdInternal_, _CMP_LT_OQ) };
348 static inline SimdDBool gmx_simdcall operator<=(SimdDouble a, SimdDouble b)
350 return { _mm512_cmp_pd_mask(a.simdInternal_, b.simdInternal_, _CMP_LE_OQ) };
353 static inline SimdDBool gmx_simdcall testBits(SimdDouble a)
355 return { _mm512_test_epi64_mask(_mm512_castpd_si512(a.simdInternal_),
356 _mm512_castpd_si512(a.simdInternal_)) };
359 static inline SimdDBool gmx_simdcall operator&&(SimdDBool a, SimdDBool b)
361 return { static_cast<__mmask8>(_mm512_kand(a.simdInternal_, b.simdInternal_)) };
364 static inline SimdDBool gmx_simdcall operator||(SimdDBool a, SimdDBool b)
366 return { static_cast<__mmask8>(_mm512_kor(a.simdInternal_, b.simdInternal_)) };
369 static inline bool gmx_simdcall anyTrue(SimdDBool a)
371 return (avx512Mask2Int(a.simdInternal_) != 0);
374 static inline SimdDouble gmx_simdcall selectByMask(SimdDouble a, SimdDBool m)
376 return { _mm512_mask_mov_pd(_mm512_setzero_pd(), m.simdInternal_, a.simdInternal_) };
379 static inline SimdDouble gmx_simdcall selectByNotMask(SimdDouble a, SimdDBool m)
381 return { _mm512_mask_mov_pd(a.simdInternal_, m.simdInternal_, _mm512_setzero_pd()) };
384 static inline SimdDouble gmx_simdcall blend(SimdDouble a, SimdDouble b, SimdDBool sel)
386 return { _mm512_mask_blend_pd(sel.simdInternal_, a.simdInternal_, b.simdInternal_) };
389 static inline SimdDouble gmx_simdcall copysign(SimdDouble a, SimdDouble b)
391 return { _mm512_castsi512_pd(_mm512_ternarylogic_epi64(_mm512_castpd_si512(a.simdInternal_),
392 _mm512_castpd_si512(b.simdInternal_),
393 _mm512_set1_epi64(INT64_MIN), 0xD8)) };
396 static inline SimdDInt32 gmx_simdcall operator&(SimdDInt32 a, SimdDInt32 b)
398 return { _mm256_and_si256(a.simdInternal_, b.simdInternal_) };
401 static inline SimdDInt32 gmx_simdcall andNot(SimdDInt32 a, SimdDInt32 b)
403 return { _mm256_andnot_si256(a.simdInternal_, b.simdInternal_) };
406 static inline SimdDInt32 gmx_simdcall operator|(SimdDInt32 a, SimdDInt32 b)
408 return { _mm256_or_si256(a.simdInternal_, b.simdInternal_) };
411 static inline SimdDInt32 gmx_simdcall operator^(SimdDInt32 a, SimdDInt32 b)
413 return { _mm256_xor_si256(a.simdInternal_, b.simdInternal_) };
416 static inline SimdDInt32 gmx_simdcall operator+(SimdDInt32 a, SimdDInt32 b)
418 return { _mm256_add_epi32(a.simdInternal_, b.simdInternal_) };
421 static inline SimdDInt32 gmx_simdcall operator-(SimdDInt32 a, SimdDInt32 b)
423 return { _mm256_sub_epi32(a.simdInternal_, b.simdInternal_) };
426 static inline SimdDInt32 gmx_simdcall operator*(SimdDInt32 a, SimdDInt32 b)
428 return { _mm256_mullo_epi32(a.simdInternal_, b.simdInternal_) };
431 static inline SimdDIBool gmx_simdcall operator==(SimdDInt32 a, SimdDInt32 b)
433 return { _mm512_mask_cmp_epi32_mask(avx512Int2Mask(0xFF), _mm512_castsi256_si512(a.simdInternal_),
434 _mm512_castsi256_si512(b.simdInternal_), _MM_CMPINT_EQ) };
437 static inline SimdDIBool gmx_simdcall testBits(SimdDInt32 a)
439 return { _mm512_mask_test_epi32_mask(avx512Int2Mask(0xFF), _mm512_castsi256_si512(a.simdInternal_),
440 _mm512_castsi256_si512(a.simdInternal_)) };
443 static inline SimdDIBool gmx_simdcall operator<(SimdDInt32 a, SimdDInt32 b)
445 return { _mm512_mask_cmp_epi32_mask(avx512Int2Mask(0xFF), _mm512_castsi256_si512(a.simdInternal_),
446 _mm512_castsi256_si512(b.simdInternal_), _MM_CMPINT_LT) };
449 static inline SimdDIBool gmx_simdcall operator&&(SimdDIBool a, SimdDIBool b)
451 return { _mm512_kand(a.simdInternal_, b.simdInternal_) };
454 static inline SimdDIBool gmx_simdcall operator||(SimdDIBool a, SimdDIBool b)
456 return { _mm512_kor(a.simdInternal_, b.simdInternal_) };
459 static inline bool gmx_simdcall anyTrue(SimdDIBool a)
461 return (avx512Mask2Int(a.simdInternal_) & 0xFF) != 0;
464 static inline SimdDInt32 gmx_simdcall selectByMask(SimdDInt32 a, SimdDIBool m)
466 return { _mm512_castsi512_si256(_mm512_mask_mov_epi32(
467 _mm512_setzero_si512(), m.simdInternal_, _mm512_castsi256_si512(a.simdInternal_))) };
470 static inline SimdDInt32 gmx_simdcall selectByNotMask(SimdDInt32 a, SimdDIBool m)
472 return { _mm512_castsi512_si256(_mm512_mask_mov_epi32(
473 _mm512_castsi256_si512(a.simdInternal_), m.simdInternal_, _mm512_setzero_si512())) };
476 static inline SimdDInt32 gmx_simdcall blend(SimdDInt32 a, SimdDInt32 b, SimdDIBool sel)
478 return { _mm512_castsi512_si256(
479 _mm512_mask_blend_epi32(sel.simdInternal_, _mm512_castsi256_si512(a.simdInternal_),
480 _mm512_castsi256_si512(b.simdInternal_))) };
483 static inline SimdDInt32 gmx_simdcall cvtR2I(SimdDouble a)
485 return { _mm512_cvtpd_epi32(a.simdInternal_) };
488 static inline SimdDInt32 gmx_simdcall cvttR2I(SimdDouble a)
490 return { _mm512_cvttpd_epi32(a.simdInternal_) };
493 static inline SimdDouble gmx_simdcall cvtI2R(SimdDInt32 a)
495 return { _mm512_cvtepi32_pd(a.simdInternal_) };
498 static inline SimdDIBool gmx_simdcall cvtB2IB(SimdDBool a)
500 return { a.simdInternal_ };
503 static inline SimdDBool gmx_simdcall cvtIB2B(SimdDIBool a)
505 return { static_cast<__mmask8>(a.simdInternal_) };
508 static inline void gmx_simdcall cvtF2DD(SimdFloat f, SimdDouble* d0, SimdDouble* d1)
510 d0->simdInternal_ = _mm512_cvtps_pd(_mm512_castps512_ps256(f.simdInternal_));
511 d1->simdInternal_ = _mm512_cvtps_pd(
512 _mm512_castps512_ps256(_mm512_shuffle_f32x4(f.simdInternal_, f.simdInternal_, 0xEE)));
515 static inline SimdFloat gmx_simdcall cvtDD2F(SimdDouble d0, SimdDouble d1)
517 __m512 f0 = _mm512_castps256_ps512(_mm512_cvtpd_ps(d0.simdInternal_));
518 __m512 f1 = _mm512_castps256_ps512(_mm512_cvtpd_ps(d1.simdInternal_));
519 return { _mm512_shuffle_f32x4(f0, f1, 0x44) };
524 #endif // GMX_SIMD_IMPL_X86_AVX_512_SIMD_DOUBLE_H