src/gromacs/simd/impl_x86_avx_512/impl_x86_avx_512_definitions.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2014,2015,2016,2017, 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.
   8  *
   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.
  13  *
  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.
  18  *
  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.
  23  *
  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.
  31  *
  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.
  34  */
  35
  36 #ifndef GMX_SIMD_IMPL_X86_AVX_512_DEFINITIONS_H
  37 #define GMX_SIMD_IMPL_X86_AVX_512_DEFINITIONS_H
  38
  39 // A couple of implementation notes:
  40 //
  41 // - We avoid using the reduce calls provided by the intel compiler, since they
  42 //   are not instructions but software routines, and not implemented on gcc.
  43 // - gcc-4.9 through 5.2 do not provide int2mask and mask2int. However, since both
  44 //   compilers are fine when we just use hexadecimal literals, we use our
  45 //   own routines avx512Mask2Int() & avx512Int2Mask() that simply return
  46 //   the argument. If some compiler needs a different solution in the future,
  47 //   just write some conditional code for these routines.
  48 // - gcc does not provide _mm512_trunc_ps(), and all alternative ops where we
  49 //   can specify a rounding mode also appear to be buggy. We conditionally work
  50 //   around this by using conversions to/from integer instead.
  51
  52 #define GMX_SIMD                                   1
  53 #define GMX_SIMD_HAVE_FLOAT                        1
  54 #define GMX_SIMD_HAVE_DOUBLE                       1
  55 #define GMX_SIMD_HAVE_LOADU                        1
  56 #define GMX_SIMD_HAVE_STOREU                       1
  57 #define GMX_SIMD_HAVE_LOGICAL                      1
  58 #define GMX_SIMD_HAVE_FMA                          1
  59 #define GMX_SIMD_HAVE_FRACTION                     0
  60 #define GMX_SIMD_HAVE_FINT32                       1
  61 #define GMX_SIMD_HAVE_FINT32_EXTRACT               0
  62 #define GMX_SIMD_HAVE_FINT32_LOGICAL               1
  63 #define GMX_SIMD_HAVE_FINT32_ARITHMETICS           1
  64 #define GMX_SIMD_HAVE_DINT32                       1
  65 // Technically it is straightforward to emulate extract on AVX-512 through
  66 // memory operations, but when applied to 16 elements as part of a table lookup
  67 // it will be faster to just store the entire vector once, so we avoid setting it.
  68 #define GMX_SIMD_HAVE_DINT32_EXTRACT               0
  69 #define GMX_SIMD_HAVE_DINT32_LOGICAL               1
  70 #define GMX_SIMD_HAVE_DINT32_ARITHMETICS           1
  71 #define GMX_SIMD_HAVE_NATIVE_COPYSIGN_FLOAT        0
  72 #define GMX_SIMD_HAVE_NATIVE_RSQRT_ITER_FLOAT      0
  73 #define GMX_SIMD_HAVE_NATIVE_RCP_ITER_FLOAT        0
  74 #define GMX_SIMD_HAVE_NATIVE_LOG_FLOAT             0
  75 #define GMX_SIMD_HAVE_NATIVE_EXP2_FLOAT            0
  76 #define GMX_SIMD_HAVE_NATIVE_EXP_FLOAT             0
  77 #define GMX_SIMD_HAVE_NATIVE_COPYSIGN_DOUBLE       0
  78 #define GMX_SIMD_HAVE_NATIVE_RSQRT_ITER_DOUBLE     0
  79 #define GMX_SIMD_HAVE_NATIVE_RCP_ITER_DOUBLE       0
  80 #define GMX_SIMD_HAVE_NATIVE_LOG_DOUBLE            0
  81 #define GMX_SIMD_HAVE_NATIVE_EXP2_DOUBLE           0
  82 #define GMX_SIMD_HAVE_NATIVE_EXP_DOUBLE            0
  83 #define GMX_SIMD_HAVE_GATHER_LOADU_BYSIMDINT_TRANSPOSE_FLOAT    1
  84 #define GMX_SIMD_HAVE_GATHER_LOADU_BYSIMDINT_TRANSPOSE_DOUBLE   1
  85 #define GMX_SIMD_HAVE_HSIMD_UTIL_FLOAT             1
  86 #define GMX_SIMD_HAVE_HSIMD_UTIL_DOUBLE            1
  87 #define GMX_SIMD_HAVE_4NSIMD_UTIL_FLOAT            1
  88 #define GMX_SIMD_HAVE_4NSIMD_UTIL_DOUBLE           1
  89
  90 #define GMX_SIMD4_HAVE_FLOAT                       1
  91 #define GMX_SIMD4_HAVE_DOUBLE                      1
  92
  93 // Implementation details
  94 #define GMX_SIMD_FLOAT_WIDTH                      16
  95 #define GMX_SIMD_DOUBLE_WIDTH                      8
  96 #define GMX_SIMD_FINT32_WIDTH                     16
  97 #define GMX_SIMD_DINT32_WIDTH                      8
  98 #define GMX_SIMD4_WIDTH                            4
  99 #define GMX_SIMD_RSQRT_BITS                       14
 100 #define GMX_SIMD_RCP_BITS                         14
 101
 102 #endif // GMX_SIMD_IMPL_X86_AVX_512_DEFINITIONS_H