2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2013,2014, 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.
37 * \defgroup module_simd SIMD intrinsics interface (simd)
38 * \ingroup group_utilitymodules
40 * \brief Provides an architecture-independent way of doing SIMD coding.
42 * Overview of the SIMD implementation is provided in \ref page_simd.
43 * The details are documented in simd.h and the reference implementation
46 * \author Erik Lindahl <erik.lindahl@scilifelab.se>
49 #ifndef GMX_SIMD_SIMD_H
50 #define GMX_SIMD_SIMD_H
52 /*! \libinternal \file
54 * \brief Definitions, capabilities, and wrappers for SIMD module.
56 * The macros in this file are intended to be used for writing
57 * architecture-independent SIMD intrinsics code.
58 * To support a new architecture, adding a new sub-include with macros here
59 * should be (nearly) all that is needed.
61 * The defines in this top-level file will set default Gromacs real precision
62 * operations to either single or double precision based on whether
63 * GMX_DOUBLE is defined. The actual implementation - including e.g.
64 * conversion operations specifically between single and double - is documented
65 * in impl_reference.h.
67 * \author Erik Lindahl <erik.lindahl@scilifelab.se>
70 * \ingroup module_simd
77 #include "gromacs/utility/basedefinitions.h"
79 /* Forward declarations so memory allocation can be used in implementations */
80 static gmx_inline float * gmx_simd_align_f(float *p);
81 static gmx_inline double * gmx_simd_align_d(double *p);
82 static gmx_inline int * gmx_simd_align_fi(int *p);
83 static gmx_inline int * gmx_simd_align_di(int *p);
84 static gmx_inline float * gmx_simd4_align_f(float *p);
85 static gmx_inline double * gmx_simd4_align_d(double *p);
88 /*! \addtogroup module_simd */
91 /*! \name SIMD predefined macros to describe high-level capabilities
93 * These macros are used to describe the features available in default
94 * Gromacs real precision. They are set from the lower-level implementation
95 * files that have macros describing single and double precision individually,
96 * as well as the implementation details.
101 * GMX_SIMD indicates that some sort of SIMD support is present in software.
103 * It is disabled if no architecture, neither reference SIMD, has been selected.
108 /* Intel MIC is a bit special since it is a co-processor. This means the rest
109 * of GROMACS (which runs on the CPU) should use a default SIMD set like AVX,
110 * while the part running on the coprocessor defines __MIC__. All functions in
111 * this SIMD module are static, so it will work perfectly fine to include this
112 * file with different SIMD definitions for different files.
115 # include "gromacs/simd/impl_intel_mic/impl_intel_mic.h"
116 #elif defined GMX_SIMD_X86_AVX2_256
117 # include "gromacs/simd/impl_x86_avx2_256/impl_x86_avx2_256.h"
118 #elif defined GMX_SIMD_X86_AVX_256
119 # include "gromacs/simd/impl_x86_avx_256/impl_x86_avx_256.h"
120 #elif defined GMX_SIMD_X86_AVX_128_FMA
121 # include "gromacs/simd/impl_x86_avx_128_fma/impl_x86_avx_128_fma.h"
122 #elif defined GMX_SIMD_X86_SSE4_1
123 # include "gromacs/simd/impl_x86_sse4_1/impl_x86_sse4_1.h"
124 #elif defined GMX_SIMD_X86_SSE2
125 # include "gromacs/simd/impl_x86_sse2/impl_x86_sse2.h"
126 #elif defined GMX_SIMD_ARM_NEON
127 # include "gromacs/simd/impl_arm_neon/impl_arm_neon.h"
128 #elif defined GMX_SIMD_IBM_QPX
129 # include "gromacs/simd/impl_ibm_qpx/impl_ibm_qpx.h"
130 #elif defined GMX_SIMD_SPARC64_HPC_ACE
131 # include "gromacs/simd/impl_sparc64_hpc_ace/impl_sparc64_hpc_ace.h"
132 #elif (defined GMX_SIMD_REFERENCE) || (defined DOXYGEN)
133 /* Plain C SIMD reference implementation, also serves as documentation.
134 * For now this code path will also be taken for Sparc64_HPC_ACE since we have
135 * not yet added the verlet kernel extensions there. The group kernels do not
136 * depend on this file, so they will still be accelerated with SIMD.
138 # include "gromacs/simd/impl_reference/impl_reference.h"
140 /* Turn off the GMX_SIMD flag if we do not even have reference support */
145 * SIMD4 width is always 4, but use this for clarity in definitions.
147 * It improves code readability to allocate e.g. 2*GMX_SIMD4_WIDTH instead of 8.
149 #define GMX_SIMD4_WIDTH 4
153 /*! \name SIMD memory alignment operations
158 * Align a float pointer for usage with SIMD instructions.
160 * You should typically \a not call this function directly (unless you explicitly
161 * want single precision even when GMX_DOUBLE is set), but use the
162 * \ref gmx_simd_align_r macro to align memory in default Gromacs real precision.
164 * \param p Pointer to memory, allocate at least \ref GMX_SIMD_FLOAT_WIDTH extra elements.
166 * \return Aligned pointer (>=p) suitable for loading/storing float fp SIMD.
167 * If \ref GMX_SIMD_HAVE_FLOAT is not set, p will be returned unchanged.
169 * Start by allocating an extra \ref GMX_SIMD_FLOAT_WIDTH float elements of memory,
170 * and then call this function. The returned pointer will be greater or equal
171 * to the one you provided, and point to an address inside your provided memory
172 * that is aligned to the SIMD width.
174 static gmx_inline float *
175 gmx_simd_align_f(float *p)
177 # ifdef GMX_SIMD_HAVE_FLOAT
178 return (float *)(((size_t)((p)+GMX_SIMD_FLOAT_WIDTH-1)) & (~((size_t)(GMX_SIMD_FLOAT_WIDTH*sizeof(float)-1))));
185 * Align a double pointer for usage with SIMD instructions.
187 * You should typically \a not call this function directly (unless you explicitly
188 * want double precision even when GMX_DOUBLE is not set), but use the
189 * \ref gmx_simd_align_r macro to align memory in default Gromacs real precision.
191 * \param p Pointer to memory, allocate at least \ref GMX_SIMD_DOUBLE_WIDTH extra elements.
193 * \return Aligned pointer (>=p) suitable for loading/storing double fp SIMD.
194 * If \ref GMX_SIMD_HAVE_DOUBLE is not set, p will be returned unchanged.
196 * Start by allocating an extra \ref GMX_SIMD_DOUBLE_WIDTH double elements of memory,
197 * and then call this function. The returned pointer will be greater or equal
198 * to the one you provided, and point to an address inside your provided memory
199 * that is aligned to the SIMD width.
201 static gmx_inline double *
202 gmx_simd_align_d(double *p)
204 # ifdef GMX_SIMD_HAVE_DOUBLE
205 return (double *)(((size_t)((p)+GMX_SIMD_DOUBLE_WIDTH-1)) & (~((size_t)(GMX_SIMD_DOUBLE_WIDTH*sizeof(double)-1))));
212 * Align a (float) integer pointer for usage with SIMD instructions.
214 * You should typically \a not call this function directly (unless you explicitly
215 * want integers corresponding to single precision even when GMX_DOUBLE is
216 * set), but use the \ref gmx_simd_align_i macro to align integer memory
217 * corresponding to Gromacs default floating-point precision.
219 * \param p Pointer to memory, allocate at least \ref GMX_SIMD_FINT32_WIDTH extra elements.
221 * \return Aligned pointer (>=p) suitable for loading/storing float-integer SIMD.
222 * If \ref GMX_SIMD_HAVE_FINT32 is not set, p will be returned unchanged.
224 * This routine provides aligned memory for usage with \ref gmx_simd_fint32_t. You
225 * should have allocated an extra \ref GMX_SIMD_FINT32_WIDTH * sizeof(int) bytes. The
226 * reason why we need to separate float-integer vs. double-integer is that the
227 * width of registers after conversions from the floating-point types might not
228 * be identical, or even supported, in both cases.
230 static gmx_inline int *
231 gmx_simd_align_fi(int *p)
233 # ifdef GMX_SIMD_HAVE_FINT32
234 return (int *)(((size_t)((p)+GMX_SIMD_FINT32_WIDTH-1)) & (~((size_t)(GMX_SIMD_FINT32_WIDTH*sizeof(int)-1))));
241 * Align a (double) integer pointer for usage with SIMD instructions.
243 * You should typically \a not call this function directly (unless you explicitly
244 * want integers corresponding to doublele precision even when GMX_DOUBLE is
245 * not set), but use the \ref gmx_simd_align_i macro to align integer memory
246 * corresponding to Gromacs default floating-point precision.
248 * \param p Pointer to memory, allocate at least \ref GMX_SIMD_DINT32_WIDTH extra elements.
250 * \return Aligned pointer (>=p) suitable for loading/storing double-integer SIMD.
251 * If \ref GMX_SIMD_HAVE_DINT32 is not set, p will be returned unchanged.
253 * This routine provides aligned memory for usage with \ref gmx_simd_dint32_t. You
254 * should have allocated an extra \ref GMX_SIMD_DINT32_WIDTH*sizeof(int) bytes. The
255 * reason why we need to separate float-integer vs. double-integer is that the
256 * width of registers after conversions from the floating-point types might not
257 * be identical, or even supported, in both cases.
259 static gmx_inline int *
260 gmx_simd_align_di(int *p)
262 # ifdef GMX_SIMD_HAVE_DINT32
263 return (int *)(((size_t)((p)+GMX_SIMD_DINT32_WIDTH-1)) & (~((size_t)(GMX_SIMD_DINT32_WIDTH*sizeof(int)-1))));
270 * Align a float pointer for usage with SIMD4 instructions.
272 * You should typically \a not call this function directly (unless you explicitly
273 * want single precision even when GMX_DOUBLE is set), but use the
274 * \ref gmx_simd4_align_r macro to align memory in default Gromacs real precision.
276 * \param p Pointer to memory, allocate at least \ref GMX_SIMD4_WIDTH extra elements.
278 * \return Aligned pointer (>=p) suitable for loading/storing float SIMD.
279 * If \ref GMX_SIMD4_HAVE_FLOAT is not set, p will be returned unchanged.
281 * This routine provides aligned memory for usage with \ref gmx_simd4_float_t.
282 * should have allocated an extra \ref GMX_SIMD4_WIDTH * sizeof(float) bytes.
284 static gmx_inline float *
285 gmx_simd4_align_f(float *p)
287 # ifdef GMX_SIMD4_HAVE_FLOAT
288 return (float *)(((size_t)((p)+GMX_SIMD4_WIDTH-1)) & (~((size_t)(GMX_SIMD4_WIDTH*sizeof(float)-1))));
295 * Align a double pointer for usage with SIMD4 instructions.
297 * You should typically \a not call this function directly (unless you explicitly
298 * want double precision even when GMX_DOUBLE is not set), but use the
299 * \ref gmx_simd4_align_r macro to align memory in default Gromacs real precision.
301 * \param p Pointer to memory, allocate at least \ref GMX_SIMD4_WIDTH extra elements.
303 * \return Aligned pointer (>=p) suitable for loading/storing float SIMD.
304 * If \ref GMX_SIMD4_HAVE_DOUBLE is not set, p will be returned unchanged.
306 * This routine provides aligned memory for usage with \ref gmx_simd4_double_t.
307 * should have allocated an extra \ref GMX_SIMD4_WIDTH * sizeof(double) bytes.
309 static gmx_inline double *
310 gmx_simd4_align_d(double *p)
312 # ifdef GMX_SIMD4_HAVE_DOUBLE
313 return (double *)(((size_t)((p)+GMX_SIMD4_WIDTH-1)) & (~((size_t)(GMX_SIMD4_WIDTH*sizeof(double)-1))));
322 /* Define Gromacs "real" precision macros depending on Gromacs config. Note
323 * that conversions float-to-double and v.v. are not included here since they
324 * are not precision-dependent - find them in the implementation files.
327 /* Double floating-point. The documentation is in the float part below */
328 # define gmx_simd_real_t gmx_simd_double_t
329 # define gmx_simd_load_r gmx_simd_load_d
330 # define gmx_simd_load1_r gmx_simd_load1_d
331 # define gmx_simd_set1_r gmx_simd_set1_d
332 # define gmx_simd_store_r gmx_simd_store_d
333 # define gmx_simd_loadu_r gmx_simd_loadu_d
334 # define gmx_simd_storeu_r gmx_simd_storeu_d
335 # define gmx_simd_setzero_r gmx_simd_setzero_d
336 # define gmx_simd_add_r gmx_simd_add_d
337 # define gmx_simd_sub_r gmx_simd_sub_d
338 # define gmx_simd_mul_r gmx_simd_mul_d
339 # define gmx_simd_fmadd_r gmx_simd_fmadd_d
340 # define gmx_simd_fmsub_r gmx_simd_fmsub_d
341 # define gmx_simd_fnmadd_r gmx_simd_fnmadd_d
342 # define gmx_simd_fnmsub_r gmx_simd_fnmsub_d
343 # define gmx_simd_and_r gmx_simd_and_d
344 # define gmx_simd_andnot_r gmx_simd_andnot_d
345 # define gmx_simd_or_r gmx_simd_or_d
346 # define gmx_simd_xor_r gmx_simd_xor_d
347 # define gmx_simd_rsqrt_r gmx_simd_rsqrt_d
348 # define gmx_simd_rcp_r gmx_simd_rcp_d
349 # define gmx_simd_fabs_r gmx_simd_fabs_d
350 # define gmx_simd_fneg_r gmx_simd_fneg_d
351 # define gmx_simd_max_r gmx_simd_max_d
352 # define gmx_simd_min_r gmx_simd_min_d
353 # define gmx_simd_round_r gmx_simd_round_d
354 # define gmx_simd_trunc_r gmx_simd_trunc_d
355 # define gmx_simd_fraction_r gmx_simd_fraction_d
356 # define gmx_simd_get_exponent_r gmx_simd_get_exponent_d
357 # define gmx_simd_get_mantissa_r gmx_simd_get_mantissa_d
358 # define gmx_simd_set_exponent_r gmx_simd_set_exponent_d
359 /* Double integer and conversions */
360 # define gmx_simd_int32_t gmx_simd_dint32_t
361 # define gmx_simd_load_i gmx_simd_load_di
362 # define gmx_simd_set1_i gmx_simd_set1_di
363 # define gmx_simd_store_i gmx_simd_store_di
364 # define gmx_simd_loadu_i gmx_simd_loadu_di
365 # define gmx_simd_storeu_i gmx_simd_storeu_di
366 # define gmx_simd_setzero_i gmx_simd_setzero_di
367 # define gmx_simd_cvt_r2i gmx_simd_cvt_d2i
368 # define gmx_simd_cvtt_r2i gmx_simd_cvtt_d2i
369 # define gmx_simd_cvt_i2r gmx_simd_cvt_i2d
370 # define gmx_simd_extract_i gmx_simd_extract_di
371 # define gmx_simd_slli_i gmx_simd_slli_di
372 # define gmx_simd_srli_i gmx_simd_srli_di
373 # define gmx_simd_and_i gmx_simd_and_di
374 # define gmx_simd_andnot_i gmx_simd_andnot_di
375 # define gmx_simd_or_i gmx_simd_or_di
376 # define gmx_simd_xor_i gmx_simd_xor_di
377 # define gmx_simd_add_i gmx_simd_add_di
378 # define gmx_simd_sub_i gmx_simd_sub_di
379 # define gmx_simd_mul_i gmx_simd_mul_di
380 /* Double booleans and selection */
381 # define gmx_simd_bool_t gmx_simd_dbool_t
382 # define gmx_simd_cmpeq_r gmx_simd_cmpeq_d
383 # define gmx_simd_cmplt_r gmx_simd_cmplt_d
384 # define gmx_simd_cmple_r gmx_simd_cmple_d
385 # define gmx_simd_and_b gmx_simd_and_db
386 # define gmx_simd_or_b gmx_simd_or_db
387 # define gmx_simd_anytrue_b gmx_simd_anytrue_db
388 # define gmx_simd_blendzero_r gmx_simd_blendzero_d
389 # define gmx_simd_blendnotzero_r gmx_simd_blendnotzero_d
390 # define gmx_simd_blendv_r gmx_simd_blendv_d
391 # define gmx_simd_reduce_r gmx_simd_reduce_d
392 # define gmx_simd_ibool_t gmx_simd_dibool_t
393 # define gmx_simd_cmpeq_i gmx_simd_cmpeq_di
394 # define gmx_simd_cmplt_i gmx_simd_cmplt_di
395 # define gmx_simd_and_ib gmx_simd_and_dib
396 # define gmx_simd_or_ib gmx_simd_or_dib
397 # define gmx_simd_anytrue_ib gmx_simd_anytrue_dib
398 # define gmx_simd_blendzero_i gmx_simd_blendzero_di
399 # define gmx_simd_blendnotzero_i gmx_simd_blendnotzero_di
400 # define gmx_simd_blendv_i gmx_simd_blendv_di
401 /* Conversions between integer and double floating-point booleans */
402 # define gmx_simd_cvt_b2ib gmx_simd_cvt_db2dib
403 # define gmx_simd_cvt_ib2b gmx_simd_cvt_dib2db
405 /* SIMD4 double fp - we only support a subset of SIMD instructions for SIMD4 */
406 # define gmx_simd4_real_t gmx_simd4_double_t
407 # define gmx_simd4_load_r gmx_simd4_load_d
408 # define gmx_simd4_load1_r gmx_simd4_load1_d
409 # define gmx_simd4_set1_r gmx_simd4_set1_d
410 # define gmx_simd4_store_r gmx_simd4_store_d
411 # define gmx_simd4_loadu_r gmx_simd4_loadu_d
412 # define gmx_simd4_storeu_r gmx_simd4_storeu_d
413 # define gmx_simd4_setzero_r gmx_simd4_setzero_d
414 # define gmx_simd4_add_r gmx_simd4_add_d
415 # define gmx_simd4_sub_r gmx_simd4_sub_d
416 # define gmx_simd4_mul_r gmx_simd4_mul_d
417 # define gmx_simd4_fmadd_r gmx_simd4_fmadd_d
418 # define gmx_simd4_fmsub_r gmx_simd4_fmsub_d
419 # define gmx_simd4_fnmadd_r gmx_simd4_fnmadd_d
420 # define gmx_simd4_fnmsub_r gmx_simd4_fnmsub_d
421 # define gmx_simd4_and_r gmx_simd4_and_d
422 # define gmx_simd4_andnot_r gmx_simd4_andnot_d
423 # define gmx_simd4_or_r gmx_simd4_or_d
424 # define gmx_simd4_xor_r gmx_simd4_xor_d
425 # define gmx_simd4_rsqrt_r gmx_simd4_rsqrt_d
426 # define gmx_simd4_fabs_r gmx_simd4_fabs_d
427 # define gmx_simd4_fneg_r gmx_simd4_fneg_d
428 # define gmx_simd4_max_r gmx_simd4_max_d
429 # define gmx_simd4_min_r gmx_simd4_min_d
430 # define gmx_simd4_round_r gmx_simd4_round_d
431 # define gmx_simd4_trunc_r gmx_simd4_trunc_d
432 # define gmx_simd4_dotproduct3_r gmx_simd4_dotproduct3_d
433 # define gmx_simd4_bool_t gmx_simd4_dbool_t
434 # define gmx_simd4_cmpeq_r gmx_simd4_cmpeq_d
435 # define gmx_simd4_cmplt_r gmx_simd4_cmplt_d
436 # define gmx_simd4_cmple_r gmx_simd4_cmple_d
437 # define gmx_simd4_and_b gmx_simd4_and_db
438 # define gmx_simd4_or_b gmx_simd4_or_db
439 # define gmx_simd4_anytrue_b gmx_simd4_anytrue_db
440 # define gmx_simd4_blendzero_r gmx_simd4_blendzero_d
441 # define gmx_simd4_blendnotzero_r gmx_simd4_blendnotzero_d
442 # define gmx_simd4_blendv_r gmx_simd4_blendv_d
443 # define gmx_simd4_reduce_r gmx_simd4_reduce_d
445 /* Memory allocation */
446 # define gmx_simd_align_r gmx_simd_align_d
447 # define gmx_simd_align_i gmx_simd_align_di
448 # define gmx_simd4_align_r gmx_simd4_align_d
450 # ifdef GMX_SIMD_HAVE_DOUBLE
451 # define GMX_SIMD_HAVE_REAL
452 # define GMX_SIMD_REAL_WIDTH GMX_SIMD_DOUBLE_WIDTH
454 # ifdef GMX_SIMD_HAVE_DINT32
455 # define GMX_SIMD_HAVE_INT32
456 # define GMX_SIMD_INT32_WIDTH GMX_SIMD_DINT32_WIDTH
458 # ifdef GMX_SIMD_HAVE_DINT32_EXTRACT
459 # define GMX_SIMD_HAVE_INT32_EXTRACT
461 # ifdef GMX_SIMD_HAVE_DINT32_LOGICAL
462 # define GMX_SIMD_HAVE_INT32_LOGICAL
464 # ifdef GMX_SIMD_HAVE_DINT32_ARITHMETICS
465 # define GMX_SIMD_HAVE_INT32_ARITHMETICS
467 # ifdef GMX_SIMD4_HAVE_DOUBLE
468 # define GMX_SIMD4_HAVE_REAL
471 #else /* GMX_DOUBLE */
473 /*! \name SIMD data types
475 * The actual storage of these types is implementation dependent. The
476 * documentation is generated from the reference implementation, but for
477 * normal usage this will likely not be what you are using.
480 /*! \brief Real precision floating-point SIMD datatype.
482 * This type is only available if \ref GMX_SIMD_HAVE_REAL is defined.
484 * If GMX_DOUBLE is defined, this will be set to \ref gmx_simd_double_t
485 * internally, otherwise \ref gmx_simd_float_t.
487 # define gmx_simd_real_t gmx_simd_float_t
489 /*! \brief 32-bit integer SIMD type.
491 * This type is only available if \ref GMX_SIMD_HAVE_INT32 is defined.
493 * If GMX_DOUBLE is defined, this will be set to \ref gmx_simd_dint32_t
494 * internally, otherwise \ref gmx_simd_fint32_t. This might seem a strange
495 * implementation detail, but it is because some SIMD implementations use
496 * different types/widths of integers registers when converting from
497 * double vs. single precision floating point. As long as you just use
498 * this type you will not have to worry about precision.
500 # define gmx_simd_int32_t gmx_simd_fint32_t
502 /*! \brief Boolean SIMD type for usage with \ref gmx_simd_real_t.
504 * This type is only available if \ref GMX_SIMD_HAVE_REAL is defined.
506 * If GMX_DOUBLE is defined, this will be set to \ref gmx_simd_dbool_t
507 * internally, otherwise \ref gmx_simd_fbool_t. This is necessary since some
508 * SIMD implementations use bitpatterns for marking truth, so single-
509 * vs. double precision booleans are not necessarily exchangable.
510 * As long as you just use this type you will not have to worry about precision.
512 * See \ref gmx_simd_ibool_t for an explanation of real vs. integer booleans.
514 # define gmx_simd_bool_t gmx_simd_fbool_t
516 /*! \brief Boolean SIMD type for usage with \ref gmx_simd_int32_t.
518 * This type is only available if \ref GMX_SIMD_HAVE_INT32 is defined.
520 * If GMX_DOUBLE is defined, this will be set to \ref gmx_simd_dibool_t
521 * internally, otherwise \ref gmx_simd_fibool_t. This is necessary since some
522 * SIMD implementations use bitpatterns for marking truth, so single-
523 * vs. double precision booleans are not necessarily exchangable, and while
524 * a double-precision boolean might be represented with a 64-bit mask, the
525 * corresponding integer might only use a 32-bit mask.
527 * We provide conversion routines for these cases, so the only thing you need to
528 * keep in mind is to use \ref gmx_simd_bool_t when working with
529 * \ref gmx_simd_real_t while you pick \ref gmx_simd_ibool_t when working with
530 * \ref gmx_simd_int32_t.
532 * To convert between them, use \ref gmx_simd_cvt_b2ib and \ref gmx_simd_cvt_ib2b.
534 # define gmx_simd_ibool_t gmx_simd_fibool_t
538 * \name SIMD load/store operations on gmx_simd_real_t
540 * \note Unaligned load/stores are only available when
541 * \ref GMX_SIMD_HAVE_LOADU and \ref GMX_SIMD_HAVE_STOREU are set, respectively.
545 /*! \brief Load \ref GMX_SIMD_REAL_WIDTH values from aligned memory to \ref gmx_simd_real_t
547 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_load_d,
548 * otherwise \ref gmx_simd_load_f.
550 * \copydetails gmx_simd_load_f
552 # define gmx_simd_load_r gmx_simd_load_f
554 /*! \brief Set all elements in \ref gmx_simd_real_t from single value in memory.
556 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_load1_d,
557 * otherwise \ref gmx_simd_load1_f.
559 * \copydetails gmx_simd_load1_f
561 # define gmx_simd_load1_r gmx_simd_load1_f
563 /*! \brief Set all elements in \ref gmx_simd_real_t from a scalar.
565 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_set1_d,
566 * otherwise \ref gmx_simd_set1_f.
568 * \copydetails gmx_simd_set1_f
570 # define gmx_simd_set1_r gmx_simd_set1_f
572 /*! \brief Store \ref GMX_SIMD_REAL_WIDTH values from \ref gmx_simd_real_t to aligned memory.
574 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_store_d,
575 * otherwise \ref gmx_simd_store_f.
577 * \copydetails gmx_simd_store_f
579 # define gmx_simd_store_r gmx_simd_store_f
581 /*! \brief Load \ref GMX_SIMD_REAL_WIDTH values from unaligned memory to \ref gmx_simd_real_t.
583 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_loadu_d,
584 * otherwise \ref gmx_simd_loadu_f.
586 * \copydetails gmx_simd_loadu_f
588 # define gmx_simd_loadu_r gmx_simd_loadu_f
590 /*! \brief Store \ref GMX_SIMD_REAL_WIDTH values from \ref gmx_simd_real_t to unaligned memory.
592 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_storeu_d,
593 * otherwise \ref gmx_simd_storeu_f.
595 * \copydetails gmx_simd_storeu_f
597 # define gmx_simd_storeu_r gmx_simd_storeu_f
599 /*! \brief Set all elements in \ref gmx_simd_real_t to 0.0.
601 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_setzero_d,
602 * otherwise \ref gmx_simd_setzero_f.
604 * \copydetails gmx_simd_setzero_f
606 # define gmx_simd_setzero_r gmx_simd_setzero_f
609 * \name SIMD load/store operations on gmx_simd_int32_t
611 * \note Unaligned load/stores are only available when
612 * \ref GMX_SIMD_HAVE_LOADU and \ref GMX_SIMD_HAVE_STOREU are set, respectively.
616 /*! \brief Load \ref GMX_SIMD_INT32_WIDTH values from aligned memory to \ref gmx_simd_int32_t .
618 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_load_di ,
619 * otherwise \ref gmx_simd_load_fi .
621 * \copydetails gmx_simd_load_fi
623 # define gmx_simd_load_i gmx_simd_load_fi
625 /*! \brief Set all elements in \ref gmx_simd_int32_t from a single integer.
627 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_set1_di ,
628 * otherwise \ref gmx_simd_set1_fi .
630 * \copydetails gmx_simd_set1_fi
632 # define gmx_simd_set1_i gmx_simd_set1_fi
634 /*! \brief Store \ref GMX_SIMD_REAL_WIDTH values from \ref gmx_simd_int32_t to aligned memory.
636 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_store_di ,
637 * otherwise \ref gmx_simd_store_fi .
639 * \copydetails gmx_simd_store_fi
641 # define gmx_simd_store_i gmx_simd_store_fi
643 /*! \brief Load \ref GMX_SIMD_REAL_WIDTH values from unaligned memory to \ref gmx_simd_int32_t.
645 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_loadu_di ,
646 * otherwise \ref gmx_simd_loadu_fi .
648 * \copydetails gmx_simd_loadu_fi
650 # define gmx_simd_loadu_i gmx_simd_loadu_fi
652 /*! \brief Store \ref GMX_SIMD_REAL_WIDTH values from \ref gmx_simd_int32_t to unaligned memory.
654 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_storeu_di ,
655 * otherwise \ref gmx_simd_storeu_fi .
657 * \copydetails gmx_simd_storeu_fi
659 # define gmx_simd_storeu_i gmx_simd_storeu_fi
661 /*! \brief Extract single integer from \ref gmx_simd_int32_t element.
663 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_extract_di ,
664 * otherwise \ref gmx_simd_extract_fi .
666 * \copydetails gmx_simd_extract_fi
668 # define gmx_simd_extract_i gmx_simd_extract_fi
670 /*! \brief Set all elements in \ref gmx_simd_int32_t to 0.
672 * If GMX_DOUBLE is defined, it will be aliased to \ref gmx_simd_setzero_di ,
673 * otherwise \ref gmx_simd_setzero_fi .
675 * \copydetails gmx_simd_setzero_fi
677 # define gmx_simd_setzero_i gmx_simd_setzero_fi
681 * \name SIMD floating-point logical operations on gmx_simd_real_t
683 * These instructions are available if \ref GMX_SIMD_HAVE_LOGICAL is defined.
687 /*! \brief Bitwise \a and on two \ref gmx_simd_real_t.
689 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_and_d,
690 * otherwise \ref gmx_simd_and_f.
692 * \copydetails gmx_simd_and_f
694 # define gmx_simd_and_r gmx_simd_and_f
696 /*! \brief Bitwise \a and-not on two \ref gmx_simd_real_t; 1st arg is complemented.
698 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_andnot_d,
699 * otherwise \ref gmx_simd_andnot_f.
701 * \copydetails gmx_simd_andnot_f
703 # define gmx_simd_andnot_r gmx_simd_andnot_f
705 /*! \brief Bitwise \a or on two \ref gmx_simd_real_t.
707 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_or_d,
708 * otherwise \ref gmx_simd_or_f.
710 * \copydetails gmx_simd_or_f
712 # define gmx_simd_or_r gmx_simd_or_f
714 /*! \brief Bitwise \a exclusive-or on two \ref gmx_simd_real_t.
716 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_xor_d,
717 * otherwise \ref gmx_simd_xor_f.
719 * \copydetails gmx_simd_xor_f
721 # define gmx_simd_xor_r gmx_simd_xor_f
724 * \name SIMD floating-point arithmetic operations on gmx_simd_real_t
728 /*! \brief SIMD a+b for two \ref gmx_simd_real_t.
730 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_add_d,
731 * otherwise \ref gmx_simd_add_f.
733 * \copydetails gmx_simd_add_f
735 # define gmx_simd_add_r gmx_simd_add_f
737 /*! \brief SIMD a-b for two \ref gmx_simd_real_t.
739 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_sub_d,
740 * otherwise \ref gmx_simd_sub_f.
742 * \copydetails gmx_simd_sub_f
744 # define gmx_simd_sub_r gmx_simd_sub_f
746 /*! \brief SIMD a*b for two \ref gmx_simd_real_t.
748 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_mul_d,
749 * otherwise \ref gmx_simd_mul_f.
751 * \copydetails gmx_simd_mul_f
753 # define gmx_simd_mul_r gmx_simd_mul_f
755 /*! \brief SIMD a*b+c for three \ref gmx_simd_real_t.
757 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fmadd_d,
758 * otherwise \ref gmx_simd_fmadd_f.
760 * \copydetails gmx_simd_fmadd_f
762 # define gmx_simd_fmadd_r gmx_simd_fmadd_f
764 /*! \brief SIMD a*b-c for three \ref gmx_simd_real_t.
766 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fmsub_d,
767 * otherwise \ref gmx_simd_fmsub_f.
769 * \copydetails gmx_simd_fmsub_f
771 # define gmx_simd_fmsub_r gmx_simd_fmsub_f
773 /*! \brief SIMD -a*b+c for three \ref gmx_simd_real_t.
775 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fnmadd_d,
776 * otherwise \ref gmx_simd_fnmadd_f.
778 * \copydetails gmx_simd_fnmadd_f
780 # define gmx_simd_fnmadd_r gmx_simd_fnmadd_f
782 /*! \brief SIMD -a*b-c for three \ref gmx_simd_real_t.
784 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fnmsub_d,
785 * otherwise \ref gmx_simd_fnmsub_f.
787 * \copydetails gmx_simd_fnmsub_f
789 # define gmx_simd_fnmsub_r gmx_simd_fnmsub_f
791 /*! \brief SIMD table lookup for 1/sqrt(x) approximation.
793 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_rsqrt_d,
794 * otherwise \ref gmx_simd_rsqrt_f.
796 * \copydetails gmx_simd_rsqrt_f
798 # define gmx_simd_rsqrt_r gmx_simd_rsqrt_f
800 /*! \brief SIMD table lookup for 1/x approximation.
802 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_rcp_d,
803 * otherwise \ref gmx_simd_rcp_f.
805 * \copydetails gmx_simd_rcp_f
807 # define gmx_simd_rcp_r gmx_simd_rcp_f
809 /*! \brief SIMD fabs(x) for \ref gmx_simd_real_t.
811 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fabs_d,
812 * otherwise \ref gmx_simd_fabs_f.
814 * \copydetails gmx_simd_fabs_f
816 # define gmx_simd_fabs_r gmx_simd_fabs_f
818 /*! \brief SIMD -x for \ref gmx_simd_real_t.
820 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fneg_d,
821 * otherwise \ref gmx_simd_fneg_f.
823 * \copydetails gmx_simd_fneg_f
825 # define gmx_simd_fneg_r gmx_simd_fneg_f
827 /*! \brief SIMD max(a,b) for each element in \ref gmx_simd_real_t.
829 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_max_d,
830 * otherwise \ref gmx_simd_max_f.
832 * \copydetails gmx_simd_max_f
834 # define gmx_simd_max_r gmx_simd_max_f
836 /*! \brief SIMD min(a,b) for each element in \ref gmx_simd_real_t.
838 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_min_d,
839 * otherwise \ref gmx_simd_min_f.
841 * \copydetails gmx_simd_min_f
843 # define gmx_simd_min_r gmx_simd_min_f
845 /*! \brief Round \ref gmx_simd_real_t to nearest int, return \ref gmx_simd_real_t.
847 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_round_d,
848 * otherwise \ref gmx_simd_round_f.
850 * \copydetails gmx_simd_round_f
852 # define gmx_simd_round_r gmx_simd_round_f
854 /*! \brief Truncate \ref gmx_simd_real_t towards 0, return \ref gmx_simd_real_t.
856 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_trunc_d,
857 * otherwise \ref gmx_simd_trunc_f.
859 * \copydetails gmx_simd_trunc_f
861 # define gmx_simd_trunc_r gmx_simd_trunc_f
863 /*! \brief SIMD Fraction, i.e. x-trunc(x) for \ref gmx_simd_real_t.
865 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_fraction_d,
866 * otherwise \ref gmx_simd_fraction_f.
868 * \copydetails gmx_simd_fraction_f
870 # define gmx_simd_fraction_r gmx_simd_fraction_f
872 /*! \brief Return the FP exponent of a SIMD \ref gmx_simd_real_t as a \ref gmx_simd_real_t.
874 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_get_exponent_d,
875 * otherwise \ref gmx_simd_get_exponent_f.
877 * \copydetails gmx_simd_exponent_f
879 # define gmx_simd_get_exponent_r gmx_simd_get_exponent_f
881 /*! \brief Return the FP mantissa of a SIMD \ref gmx_simd_real_t as a \ref gmx_simd_real_t.
883 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_get_mantissa_d,
884 * otherwise \ref gmx_simd_get_mantissa_f.
886 * \copydetails gmx_simd_mantissa_f
888 # define gmx_simd_get_mantissa_r gmx_simd_get_mantissa_f
890 /*! \brief Set the exponent of a SIMD \ref gmx_simd_real_t from a \ref gmx_simd_real_t.
892 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_set_exponent_d,
893 * otherwise \ref gmx_simd_set_exponent_f.
895 * \copydetails gmx_simd_set_exponent_f
897 # define gmx_simd_set_exponent_r gmx_simd_set_exponent_f
900 * \name SIMD comparison, boolean, and select operations for gmx_simd_real_t
904 /*! \brief SIMD a==b for \ref gmx_simd_real_t. Returns a \ref gmx_simd_bool_t.
906 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cmpeq_d,
907 * otherwise \ref gmx_simd_cmpeq_f.
909 * \copydetails gmx_simd_cmpeq_f
911 # define gmx_simd_cmpeq_r gmx_simd_cmpeq_f
913 /*! \brief SIMD a<b for \ref gmx_simd_real_t. Returns a \ref gmx_simd_bool_t.
915 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cmplt_d,
916 * otherwise \ref gmx_simd_cmplt_f.
918 * \copydetails gmx_simd_cmplt_f
920 # define gmx_simd_cmplt_r gmx_simd_cmplt_f
922 /*! \brief SIMD a<=b for \ref gmx_simd_real_t. Returns a \ref gmx_simd_bool_t.
924 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cmple_d,
925 * otherwise \ref gmx_simd_cmple_f.
927 * \copydetails gmx_simd_cmple_f
929 # define gmx_simd_cmple_r gmx_simd_cmple_f
931 /*! \brief For each element, the result boolean is true if both arguments are true
933 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_and_db,
934 * otherwise \ref gmx_simd_and_fb.
936 * \copydetails gmx_simd_and_fb
938 # define gmx_simd_and_b gmx_simd_and_fb
940 /*! \brief For each element, the result boolean is true if either argument is true
942 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_or_db,
943 * otherwise \ref gmx_simd_or_fb.
945 * \copydetails gmx_simd_or_fn
947 # define gmx_simd_or_b gmx_simd_or_fb
949 /*! \brief Return nonzero if any element in gmx_simd_bool_t is true, otherwise 0.
951 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_anytrue_db,
952 * otherwise \ref gmx_simd_anytrue_fb.
954 * \copydetails gmx_simd_anytrue_fb
956 # define gmx_simd_anytrue_b gmx_simd_anytrue_fb
958 /*! \brief Selects elements from \ref gmx_simd_real_t where boolean is true, otherwise 0.
960 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendzero_d,
961 * otherwise \ref gmx_simd_blendzero_f.
963 * \copydetails gmx_simd_blendzero_f
965 * \sa gmx_simd_blendzero_i
967 # define gmx_simd_blendzero_r gmx_simd_blendzero_f
969 /*! \brief Selects elements from \ref gmx_simd_real_t where boolean is false, otherwise 0.
971 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendnotzero_d,
972 * otherwise \ref gmx_simd_blendnotzero_f.
974 * \copydetails gmx_simd_blendnotzero_f
976 # define gmx_simd_blendnotzero_r gmx_simd_blendnotzero_f
978 /*! \brief Selects from 2nd real SIMD arg where boolean is true, otherwise 1st arg.
980 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendv_d,
981 * otherwise \ref gmx_simd_blendv_f.
983 * \copydetails gmx_simd_blendv_f
985 # define gmx_simd_blendv_r gmx_simd_blendv_f
987 /*! \brief Return sum of all elements in SIMD floating-point variable.
989 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_reduce_d,
990 * otherwise \ref gmx_simd_reduce_f.
992 * \copydetails gmx_simd_reduce_f
994 # define gmx_simd_reduce_r gmx_simd_reduce_f
997 * \name SIMD integer logical operations on gmx_simd_int32_t
999 * These instructions are available if \ref GMX_SIMD_HAVE_INT32_LOGICAL is defined.
1003 /*! \brief Shift each element in \ref gmx_simd_int32_t left by immediate
1005 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_slli_di,
1006 * otherwise \ref gmx_simd_slli_fi.
1008 * \copydetails gmx_simd_slli_fi
1010 # define gmx_simd_slli_i gmx_simd_slli_fi
1012 /*! \brief Shift each element in \ref gmx_simd_int32_t right by immediate
1014 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_srli_di,
1015 * otherwise \ref gmx_simd_srli_fi.
1017 * \copydetails gmx_simd_srli_fi
1019 # define gmx_simd_srli_i gmx_simd_srli_fi
1021 /*! \brief Bitwise \a and on two \ref gmx_simd_int32_t.
1023 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_and_di,
1024 * otherwise \ref gmx_simd_and_fi.
1026 * \copydetails gmx_simd_and_fi
1028 # define gmx_simd_and_i gmx_simd_and_fi
1030 /*! \brief Bitwise \a and-not on two \ref gmx_simd_int32_t; 1st arg is complemented.
1032 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_andnot_di,
1033 * otherwise \ref gmx_simd_andnot_fi.
1035 * \copydetails gmx_simd_andnot_fi
1037 # define gmx_simd_andnot_i gmx_simd_andnot_fi
1039 /*! \brief Bitwise \a or on two \ref gmx_simd_int32_t.
1041 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_or_di,
1042 * otherwise \ref gmx_simd_or_fi.
1044 * \copydetails gmx_simd_or_fi
1046 # define gmx_simd_or_i gmx_simd_or_fi
1048 /*! \brief Bitwise \a xor on two \ref gmx_simd_int32_t.
1050 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_xor_di,
1051 * otherwise \ref gmx_simd_xor_fi.
1053 * \copydetails gmx_simd_xor_fi
1055 # define gmx_simd_xor_i gmx_simd_xor_fi
1058 * \name SIMD integer arithmetic operations on gmx_simd_int32_t
1060 * These instructions are available if \ref GMX_SIMD_HAVE_INT32_ARITHMETICS is defined.
1064 /*! \brief SIMD a+b for two \ref gmx_simd_int32_t.
1066 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_add_di,
1067 * otherwise \ref gmx_simd_add_fi.
1069 * \copydetails gmx_simd_add_fi
1071 # define gmx_simd_add_i gmx_simd_add_fi
1073 /*! \brief SIMD a-b for two \ref gmx_simd_int32_t.
1075 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_sub_di,
1076 * otherwise \ref gmx_simd_sub_fi.
1078 * \copydetails gmx_simd_sub_fi
1080 # define gmx_simd_sub_i gmx_simd_sub_fi
1082 /*! \brief SIMD a*b for two \ref gmx_simd_int32_t.
1084 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_mul_di,
1085 * otherwise \ref gmx_simd_mul_fi.
1087 * \copydetails gmx_simd_mul_fi
1089 # define gmx_simd_mul_i gmx_simd_mul_fi
1092 * \name SIMD integer comparison, booleans, and selection on gmx_simd_int32_t
1094 * These instructions are available if \ref GMX_SIMD_HAVE_INT32_ARITHMETICS is defined.
1098 /*! \brief Returns boolean describing whether a==b, for \ref gmx_simd_int32_t
1100 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cmpeq_di,
1101 * otherwise \ref gmx_simd_cmpeq_fi.
1103 * \copydetails gmx_simd_cmpeq_fi
1105 # define gmx_simd_cmpeq_i gmx_simd_cmpeq_fi
1107 /*! \brief Returns boolean describing whether a<b, for \ref gmx_simd_int32_t
1109 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cmplt_di,
1110 * otherwise \ref gmx_simd_cmplt_fi.
1112 * \copydetails gmx_simd_cmplt_fi
1114 # define gmx_simd_cmplt_i gmx_simd_cmplt_fi
1116 /*! \brief For each element, the result boolean is true if both arguments are true
1118 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_and_dib,
1119 * otherwise \ref gmx_simd_and_fib.
1121 * \copydetails gmx_simd_and_fib
1123 # define gmx_simd_and_ib gmx_simd_and_fib
1125 /*! \brief For each element, the result boolean is true if either argument is true.
1127 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_or_dib,
1128 * otherwise \ref gmx_simd_or_fib.
1130 * \copydetails gmx_simd_or_fib
1132 # define gmx_simd_or_ib gmx_simd_or_fib
1134 /*! \brief Return nonzero if any element in gmx_simd_ibool_t is true, otherwise 0.
1136 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_anytrue_dib,
1137 * otherwise \ref gmx_simd_anytrue_fib.
1139 * \copydetails gmx_simd_anytrue_fib
1141 # define gmx_simd_anytrue_ib gmx_simd_anytrue_fib
1143 /*! \brief Selects elements from \ref gmx_simd_int32_t where boolean is true, otherwise 0.
1145 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendzero_di,
1146 * otherwise \ref gmx_simd_blendzero_fi.
1148 * \copydetails gmx_simd_blendzero_fi
1150 # define gmx_simd_blendzero_i gmx_simd_blendzero_fi
1152 /*! \brief Selects elements from \ref gmx_simd_int32_t where boolean is false, otherwise 0.
1154 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendnotzero_di,
1155 * otherwise \ref gmx_simd_blendnotzero_fi.
1157 * \copydetails gmx_simd_blendnotzero_fi
1159 # define gmx_simd_blendnotzero_i gmx_simd_blendnotzero_fi
1161 /*! \brief Selects from 2nd int SIMD arg where boolean is true, otherwise 1st arg.
1163 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_blendv_di,
1164 * otherwise \ref gmx_simd_blendv_fi.
1166 * \copydetails gmx_simd_blendv_fi
1168 # define gmx_simd_blendv_i gmx_simd_blendv_fi
1171 * \name SIMD conversion operations
1173 * These instructions are available when both types involved in the conversion
1174 * are defined, e.g. \ref GMX_SIMD_HAVE_REAL and \ref GMX_SIMD_HAVE_INT32
1175 * for real-to-integer conversion.
1179 /*! \brief Convert gmx_simd_real_t to gmx_simd_int32_t, round to nearest integer.
1181 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cvt_d2i,
1182 * otherwise \ref gmx_simd_cvt_f2i.
1184 * \copydetails gmx_simd_cvt_f2i
1186 # define gmx_simd_cvt_r2i gmx_simd_cvt_f2i
1188 /*! \brief Convert gmx_simd_real_t to gmx_simd_int32_t, truncate towards zero
1190 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cvtt_d2i,
1191 * otherwise \ref gmx_simd_cvtt_f2i.
1193 * \copydetails gmx_simd_cvtt_f2i
1195 # define gmx_simd_cvtt_r2i gmx_simd_cvtt_f2i
1197 /*! \brief Convert gmx_simd_int32_t to gmx_simd_real_t
1199 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cvt_i2d,
1200 * otherwise \ref gmx_simd_cvt_i2f.
1202 * \copydetails gmx_simd_cvt_i2f
1204 # define gmx_simd_cvt_i2r gmx_simd_cvt_i2f
1206 /*! \brief Convert from gmx_simd_bool_t to gmx_simd_ibool_t
1208 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cvt_db2dib,
1209 * otherwise \ref gmx_simd_cvt_fb2fib.
1211 * \copydetails gmx_simd_cvt_fb2fib
1213 # define gmx_simd_cvt_b2ib gmx_simd_cvt_fb2fib
1215 /*! \brief Convert from gmx_simd_ibool_t to gmx_simd_bool_t
1217 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_cvt_dib2db,
1218 * otherwise \ref gmx_simd_cvt_fib2fb.
1220 * \copydetails gmx_simd_cvt_fib2fb
1222 # define gmx_simd_cvt_ib2b gmx_simd_cvt_fib2fb
1226 * \name SIMD memory alignment operations
1230 /*! \brief Align real memory for SIMD usage.
1232 * This routine will only align memory if \ref GMX_SIMD_HAVE_REAL is defined.
1233 * Otherwise the original pointer will be returned.
1235 * Start by allocating an extra \ref GMX_SIMD_REAL_WIDTH float elements of memory,
1236 * and then call this function. The returned pointer will be greater or equal
1237 * to the one you provided, and point to an address inside your provided memory
1238 * that is aligned to the SIMD width.
1240 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_align_d,
1241 * otherwise \ref gmx_simd_align_f. For detailed documentation, see the
1242 * precision-specific implementation routines.
1244 # define gmx_simd_align_r gmx_simd_align_f
1246 /*! \brief Align integer memory for SIMD usage.
1248 * This routine will only align memory if \ref GMX_SIMD_HAVE_INT32 is defined.
1249 * Otherwise the original pointer will be returned.
1251 * Start by allocating an extra \ref GMX_SIMD_INT32_WIDTH elements of memory,
1252 * and then call this function. The returned pointer will be greater or equal
1253 * to the one you provided, and point to an address inside your provided memory
1254 * that is aligned to the SIMD width.
1256 * If GMX_DOUBLE is defined, this will be aliased to \ref gmx_simd_align_di,
1257 * otherwise \ref gmx_simd_align_fi. For detailed documentation, see the
1258 * precision-specific implementation routines.
1260 # define gmx_simd_align_i gmx_simd_align_fi
1264 /*! \name SIMD4 - constant width-four SIMD datatypes
1266 * These operations are only meant to be used for a few coordinate
1267 * manipulation and grid interpolation routines, so we only support a subset
1268 * of operations for SIMD4. To avoid repeating all the documentation from
1269 * the generic width SIMD routines, we only provide brief documentation for
1270 * these operations. Follow the link to the implementation documentation or the
1271 * reference to the corresponding generic SIMD routine. The format will be
1272 * exactly the same, but they have SIMD replaced with SIMD4.
1276 /*! \brief SIMD real datatype guaranteed to be 4 elements wide, if available.
1278 * All the SIMD4 datatypes and operations behave like their counterparts for
1279 * the generic SIMD implementation, but they might be implemented with different
1280 * registers, or not supported at all. It is important that you check the
1281 * define \ref GMX_SIMD4_HAVE_REAL before using it.
1283 * Just as the normal SIMD operations, all SIMD4 types and routines will
1284 * be aliased to either single or double precision ones based on whether
1285 * GMX_DOUBLE is defined.
1287 * \note There is no support for integer or math operations in SIMD4.
1289 # define gmx_simd4_real_t gmx_simd4_float_t
1291 /*! \brief Boolean for \ref gmx_simd4_real_t comparision/selection */
1292 # define gmx_simd4_bool_t gmx_simd4_fbool_t
1294 /*! \brief Load aligned data to gmx_simd4_real_t.
1296 * \copydetails gmx_simd4_load_f
1298 # define gmx_simd4_load_r gmx_simd4_load_f
1300 /*! \brief Load single element to gmx_simd4_real_t
1302 * \copydetails gmx_simd4_load1_f
1304 # define gmx_simd4_load1_r gmx_simd4_load1_f
1306 /*! \brief Set gmx_simd4_real_t from scalar value
1308 * \copydetails gmx_simd4_set1_f
1310 # define gmx_simd4_set1_r gmx_simd4_set1_f
1312 /*! \brief store aligned data from gmx_simd4_real_t
1314 * \copydetails gmx_simd4_store_f
1316 # define gmx_simd4_store_r gmx_simd4_store_f
1318 /*! \brief Load unaligned data to gmx_simd4_real_t
1320 * \copydetails gmx_simd4_loadu_f
1322 # define gmx_simd4_loadu_r gmx_simd4_loadu_f
1324 /*! \brief Store unaligned data from gmx_simd4_real_t
1326 * \copydetails gmx_simd4_storeu_f
1328 # define gmx_simd4_storeu_r gmx_simd4_storeu_f
1330 /*! \brief Set all elements in gmx_simd4_real_t to 0.0
1332 * \copydetails gmx_simd4_setzero_f
1334 # define gmx_simd4_setzero_r gmx_simd4_setzero_f
1336 /*! \brief Bitwise and for two gmx_simd4_real_t
1338 * \copydetails gmx_simd4_and_f
1340 # define gmx_simd4_and_r gmx_simd4_and_f
1342 /*! \brief Bitwise and-not for two gmx_simd4_real_t. 1st arg is complemented.
1344 * \copydetails gmx_simd4_andnot_f
1346 # define gmx_simd4_andnot_r gmx_simd4_andnot_f
1348 /*! \brief Bitwise or for two gmx_simd4_real_t
1350 * \copydetails gmx_simd4_or_f
1352 # define gmx_simd4_or_r gmx_simd4_or_f
1354 /*! \brief Bitwise xor for two gmx_simd4_real_t
1356 * \copydetails gmx_simd4_xor_f
1358 # define gmx_simd4_xor_r gmx_simd4_xor_f
1360 /*! \brief a+b for \ref gmx_simd4_real_t
1362 * \copydetails gmx_simd4_add_f
1364 # define gmx_simd4_add_r gmx_simd4_add_f
1366 /*! \brief a-b for \ref gmx_simd4_real_t
1368 * \copydetails gmx_simd4_sub_f
1370 # define gmx_simd4_sub_r gmx_simd4_sub_f
1372 /*! \brief a*b for \ref gmx_simd4_real_t
1374 * \copydetails gmx_simd4_mul_f
1376 # define gmx_simd4_mul_r gmx_simd4_mul_f
1378 /*! \brief a*b+c for \ref gmx_simd4_real_t
1380 * \copydetails gmx_simd4_fmadd_f
1382 # define gmx_simd4_fmadd_r gmx_simd4_fmadd_f
1384 /*! \brief a*b-c for \ref gmx_simd4_real_t
1386 * \copydetails gmx_simd4_fmsub_f
1388 # define gmx_simd4_fmsub_r gmx_simd4_fmsub_f
1390 /*! \brief -a*b+c for \ref gmx_simd4_real_t
1392 * \copydetails gmx_simd4_fnmadd_f
1394 # define gmx_simd4_fnmadd_r gmx_simd4_fnmadd_f
1396 /*! \brief -a*b-c for \ref gmx_simd4_real_t
1398 * \copydetails gmx_simd4_fnmsub_f
1400 # define gmx_simd4_fnmsub_r gmx_simd4_fnmsub_f
1402 /*! \brief 1/sqrt(x) approximate lookup for \ref gmx_simd4_real_t
1404 * \copydetails gmx_simd4_rsqrt_f
1406 # define gmx_simd4_rsqrt_r gmx_simd4_rsqrt_f
1408 /*! \brief fabs(x) for \ref gmx_simd4_real_t
1410 * \copydetails gmx_simd4_fabs_f
1412 # define gmx_simd4_fabs_r gmx_simd4_fabs_f
1414 /*! \brief Change sign (-x) for \ref gmx_simd4_real_t
1416 * \copydetails gmx_simd4_fneg_f
1418 # define gmx_simd4_fneg_r gmx_simd4_fneg_f
1420 /*! \brief Select maximum of each pair of elements from args for \ref gmx_simd4_real_t
1422 * \copydetails gmx_simd4_max_f
1424 # define gmx_simd4_max_r gmx_simd4_max_f
1426 /*! \brief Select minimum of each pair of elements from args for \ref gmx_simd4_real_t
1428 * \copydetails gmx_simd4_min_f
1430 # define gmx_simd4_min_r gmx_simd4_min_f
1432 /*! \brief Round \ref gmx_simd4_real_t to nearest integer, return \ref gmx_simd4_real_t
1434 * \copydetails gmx_simd4_round_f
1436 # define gmx_simd4_round_r gmx_simd4_round_f
1438 /*! \brief Truncate \ref gmx_simd4_real_t towards zero, return \ref gmx_simd4_real_t
1440 * \copydetails gmx_simd4_trunc_f
1442 # define gmx_simd4_trunc_r gmx_simd4_trunc_f
1444 /*! \brief Scalar product of first three elements of two \ref gmx_simd4_real_t *
1446 * \copydetails gmx_simd4_dotproduct3_f
1448 # define gmx_simd4_dotproduct3_r gmx_simd4_dotproduct3_f
1450 /*! \brief Return booleans whether a==b for each element two \ref gmx_simd4_real_t
1452 * \copydetails gmx_simd4_cmpeq_f
1454 # define gmx_simd4_cmpeq_r gmx_simd4_cmpeq_f
1455 /*! \brief Return booleans whether a<b for each element two \ref gmx_simd4_real_t
1457 * \copydetails gmx_simd4_cmplt_f
1459 # define gmx_simd4_cmplt_r gmx_simd4_cmplt_f
1460 /*! \brief Return booleans whether a<=b for each element two \ref gmx_simd4_real_t
1462 * \copydetails gmx_simd4_cmple_f
1464 # define gmx_simd4_cmple_r gmx_simd4_cmple_f
1466 /*! \brief Logical and for two \ref gmx_simd4_bool_t
1468 * \copydetails gmx_simd4_and_fb
1470 # define gmx_simd4_and_b gmx_simd4_and_fb
1471 /*! \brief Logical or for two \ref gmx_simd4_bool_t
1473 * \copydetails gmx_simd4_or_fb
1475 # define gmx_simd4_or_b gmx_simd4_or_fb
1477 /*! \brief Return nonzero if any element in \ref gmx_simd4_bool_t is true, otherwise 0
1479 * \copydetails gmx_simd4_anytrue_fb
1481 # define gmx_simd4_anytrue_b gmx_simd4_anytrue_fb
1483 /*! \brief Selects from 2nd real SIMD4 arg where boolean is true, otherwise 1st arg
1485 * \copydetails gmx_simd4_blendzero_f
1487 # define gmx_simd4_blendzero_r gmx_simd4_blendzero_f
1489 /*! \brief Selects from 2nd real SIMD4 arg where boolean is false, otherwise 1st arg
1491 * \copydetails gmx_simd4_blendnotzero_f
1493 # define gmx_simd4_blendnotzero_r gmx_simd4_blendnotzero_f
1495 /*! \brief Selects from 2nd real SIMD4 arg where boolean is true, otherwise 1st arg
1497 * \copydetails gmx_simd4_blendv_f
1499 # define gmx_simd4_blendv_r gmx_simd4_blendv_f
1501 /*! \brief Return sum of all elements in SIMD4 floating-point variable.
1503 * \copydetails gmx_simd4_reduce_f
1505 # define gmx_simd4_reduce_r gmx_simd4_reduce_f
1507 /*! \brief Align real memory for SIMD4 usage.
1509 * \copydetails gmx_simd4_align_f
1511 # define gmx_simd4_align_r gmx_simd4_align_f
1515 /*! \name SIMD predefined macros to describe high-level capabilities
1519 # if (defined GMX_SIMD_HAVE_FLOAT) || (defined DOXYGEN)
1520 /*! \brief Defined if gmx_simd_real_t is available.
1522 * if GMX_DOUBLE is defined, this will be aliased to
1523 * \ref GMX_SIMD_HAVE_DOUBLE, otherwise GMX_SIMD_HAVE_FLOAT.
1525 # define GMX_SIMD_HAVE_REAL
1526 /*! \brief Width of gmx_simd_real_t.
1528 * if GMX_DOUBLE is defined, this will be aliased to
1529 * \ref GMX_SIMD_DOUBLE_WIDTH, otherwise GMX_SIMD_FLOAT_WIDTH.
1531 # define GMX_SIMD_REAL_WIDTH GMX_SIMD_FLOAT_WIDTH
1533 # if (defined GMX_SIMD_HAVE_FINT32) || (defined DOXYGEN)
1534 /*! \brief Defined if gmx_simd_int32_t is available.
1536 * if GMX_DOUBLE is defined, this will be aliased to
1537 * \ref GMX_SIMD_HAVE_DINT32, otherwise GMX_SIMD_HAVE_FINT32.
1539 # define GMX_SIMD_HAVE_INT32
1540 /*! \brief Width of gmx_simd_int32_t.
1542 * if GMX_DOUBLE is defined, this will be aliased to
1543 * \ref GMX_SIMD_DINT32_WIDTH, otherwise GMX_SIMD_FINT32_WIDTH.
1545 # define GMX_SIMD_INT32_WIDTH GMX_SIMD_FINT32_WIDTH
1547 # if (defined GMX_SIMD_HAVE_FINT32_EXTRACT) || (defined DOXYGEN)
1548 /*! \brief Defined if gmx_simd_extract_i() is available.
1550 * if GMX_DOUBLE is defined, this will be aliased to
1551 * \ref GMX_SIMD_HAVE_DINT32_EXTRACT, otherwise GMX_SIMD_HAVE_FINT32_EXTRACT.
1553 # define GMX_SIMD_HAVE_INT32_EXTRACT
1555 # if (defined GMX_SIMD_HAVE_FINT32_LOGICAL) || (defined DOXYGEN)
1556 /*! \brief Defined if logical ops are supported on gmx_simd_int32_t.
1558 * if GMX_DOUBLE is defined, this will be aliased to
1559 * \ref GMX_SIMD_HAVE_DINT32_LOGICAL, otherwise GMX_SIMD_HAVE_FINT32_LOGICAL.
1561 # define GMX_SIMD_HAVE_INT32_LOGICAL
1563 # if (defined GMX_SIMD_HAVE_FINT32_ARITHMETICS) || (defined DOXYGEN)
1564 /*! \brief Defined if arithmetic ops are supported on gmx_simd_int32_t.
1566 * if GMX_DOUBLE is defined, this will be aliased to
1567 * \ref GMX_SIMD_HAVE_DINT32_ARITHMETICS, otherwise GMX_SIMD_HAVE_FINT32_ARITHMETICS.
1569 # define GMX_SIMD_HAVE_INT32_ARITHMETICS
1571 # if (defined GMX_SIMD4_HAVE_FLOAT) || (defined DOXYGEN)
1572 /*! \brief Defined if gmx_simd4_real_t is available.
1574 * if GMX_DOUBLE is defined, this will be aliased to
1575 * \ref GMX_SIMD4_HAVE_DOUBLE, otherwise GMX_SIMD4_HAVE_FLOAT.
1577 # define GMX_SIMD4_HAVE_REAL
1582 #endif /* GMX_DOUBLE */
1587 #endif /* GMX_SIMD_SIMD_H */