2 * This file is part of the GROMACS molecular simulation package.
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.
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_KNL_SIMD_FLOAT_H
37 #define GMX_SIMD_IMPL_X86_AVX_512_KNL_SIMD_FLOAT_H
41 #include <immintrin.h>
43 #include "gromacs/math/utilities.h"
44 #include "gromacs/simd/impl_x86_avx_512/impl_x86_avx_512_simd_float.h"
49 static inline SimdFloat gmx_simdcall
53 _mm512_rsqrt28_ps(x.simdInternal_)
57 static inline SimdFloat gmx_simdcall
61 _mm512_rcp28_ps(x.simdInternal_)
65 static inline SimdFloat gmx_simdcall
66 maskzRsqrt(SimdFloat x, SimdFBool m)
69 _mm512_maskz_rsqrt28_ps(m.simdInternal_, x.simdInternal_)
73 static inline SimdFloat gmx_simdcall
74 maskzRcp(SimdFloat x, SimdFBool m)
77 _mm512_maskz_rcp28_ps(m.simdInternal_, x.simdInternal_)
81 template <MathOptimization opt = MathOptimization::Safe>
82 static inline SimdFloat gmx_simdcall
86 _mm512_exp2a23_ps(x.simdInternal_)
90 template <MathOptimization opt = MathOptimization::Safe>
91 static inline SimdFloat gmx_simdcall
94 const __m512 argscale = _mm512_set1_ps(1.44269504088896341f);
95 const __m512 invargscale = _mm512_set1_ps(-0.69314718055994528623f);
97 if (opt == MathOptimization::Safe)
99 // Set the limit to gurantee flush to zero
100 const SimdFloat smallArgLimit(-88.f);
101 // Since we multiply the argument by 1.44, for the safe version we need to make
102 // sure this doesn't result in overflow
103 x = max(x, smallArgLimit);
106 __m512 xscaled = _mm512_mul_ps(x.simdInternal_, argscale);
107 __m512 r = _mm512_exp2a23_ps(xscaled);
109 // exp2a23_ps provides 23 bits of accuracy, but we ruin some of that with our argument
110 // scaling. To correct this, we find the difference between the scaled argument and
111 // the true one (extended precision arithmetics does not appear to be necessary to
112 // fulfill our accuracy requirements) and then multiply by the exponent of this
113 // correction since exp(a+b)=exp(a)*exp(b).
114 // Note that this only adds two instructions (and maybe some constant loads).
116 // find the difference
117 x = _mm512_fmadd_ps(invargscale, xscaled, x.simdInternal_);
118 // x will now be a _very_ small number, so approximate exp(x)=1+x.
119 // We should thus apply the correction as r'=r*(1+x)=r+r*x
120 r = _mm512_fmadd_ps(r, x.simdInternal_, r);
128 #endif // GMX_SIMD_IMPL_X86_AVX_512_KNL_SIMD_FLOAT_H