2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012, by the GROMACS development team, led by
5 * David van der Spoel, Berk Hess, Erik Lindahl, and including many
6 * others, as listed in the AUTHORS file in the top-level source
7 * 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.
35 #ifndef _gmx_x86_avx_128_fma_h_
36 #define _gmx_x86_avx_128_fma_h_
39 #include <immintrin.h>
40 #ifdef HAVE_X86INTRIN_H
41 #include <x86intrin.h> /* FMA */
46 #include "types/simple.h"
49 #define gmx_mm_extract_epi32(x, imm) _mm_cvtsi128_si32(_mm_srli_si128((x), 4 * (imm)))
51 #define _GMX_MM_BLEND(b3,b2,b1,b0) (((b3) << 3) | ((b2) << 2) | ((b1) << 1) | ((b0)))
53 #define _GMX_MM_PERMUTE128D(fp1,fp0) (((fp1) << 1) | ((fp0)))
56 #define GMX_MM_TRANSPOSE2_PD(row0, row1) { \
57 __m128d __gmx_t1 = row0; \
58 row0 = _mm_unpacklo_pd(row0,row1); \
59 row1 = _mm_unpackhi_pd(__gmx_t1,row1); \
63 #if (defined (_MSC_VER) || defined(__INTEL_COMPILER))
64 # define gmx_mm_castsi128_ps(a) _mm_castsi128_ps(a)
65 # define gmx_mm_castps_si128(a) _mm_castps_si128(a)
66 # define gmx_mm_castps_ps128(a) (a)
67 # define gmx_mm_castsi128_pd(a) _mm_castsi128_pd(a)
68 # define gmx_mm_castpd_si128(a) _mm_castpd_si128(a)
69 #elif defined(__GNUC__)
70 # define gmx_mm_castsi128_ps(a) ((__m128)(a))
71 # define gmx_mm_castps_si128(a) ((__m128i)(a))
72 # define gmx_mm_castps_ps128(a) ((__m128)(a))
73 # define gmx_mm_castsi128_pd(a) ((__m128d)(a))
74 # define gmx_mm_castpd_si128(a) ((__m128i)(a))
76 static __m128 gmx_mm_castsi128_ps(__m128i a)
78 return *(__m128 *) &a;
80 static __m128i gmx_mm_castps_si128(__m128 a)
82 return *(__m128i *) &a;
84 static __m128 gmx_mm_castps_ps128(__m128 a)
86 return *(__m128 *) &a;
88 static __m128d gmx_mm_castsi128_pd(__m128i a)
90 return *(__m128d *) &a;
92 static __m128i gmx_mm_castpd_si128(__m128d a)
94 return *(__m128i *) &a;
98 #if GMX_EMULATE_AMD_FMA
99 /* Wrapper routines so we can do test builds on non-FMA or non-AMD hardware */
101 _mm_macc_ps(__m128 a, __m128 b, __m128 c)
103 return _mm_add_ps(c,_mm_mul_ps(a,b));
107 _mm_nmacc_ps(__m128 a, __m128 b, __m128 c)
109 return _mm_sub_ps(c,_mm_mul_ps(a,b));
113 _mm_msub_ps(__m128 a, __m128 b, __m128 c)
115 return _mm_sub_ps(_mm_mul_ps(a,b),c);
119 _mm_macc_pd(__m128d a, __m128d b, __m128d c)
121 return _mm_add_pd(c,_mm_mul_pd(a,b));
125 _mm_nmacc_pd(__m128d a, __m128d b, __m128d c)
127 return _mm_sub_pd(c,_mm_mul_pd(a,b));
131 _mm_msub_pd(__m128d a, __m128d b, __m128d c)
133 return _mm_sub_pd(_mm_mul_pd(a,b),c);
135 #endif /* AMD FMA emulation support */
138 gmx_mm_printxmm_ps(const char *s,__m128 xmm)
142 _mm_storeu_ps(f,xmm);
143 printf("%s: %15.10e %15.10e %15.10e %15.10e\n",s,f[0],f[1],f[2],f[3]);
148 gmx_mm_printxmmsum_ps(const char *s,__m128 xmm)
152 _mm_storeu_ps(f,xmm);
153 printf("%s (sum): %15.10g\n",s,f[0]+f[1]+f[2]+f[3]);
158 gmx_mm_printxmm_pd(const char *s,__m128d xmm)
162 _mm_storeu_pd(f,xmm);
163 printf("%s: %30.20e %30.20e\n",s,f[0],f[1]);
167 gmx_mm_printxmmsum_pd(const char *s,__m128d xmm)
171 _mm_storeu_pd(f,xmm);
172 printf("%s (sum): %15.10g\n",s,f[0]+f[1]);
177 gmx_mm_printxmm_epi32(const char *s,__m128i xmmi)
181 _mm_storeu_si128((__m128i *)i,xmmi);
182 printf("%10s: %2d %2d %2d %2d\n",s,i[0],i[1],i[2],i[3]);
187 static int gmx_mm_check_and_reset_overflow(void)
192 MXCSR = _mm_getcsr();
193 /* The overflow flag is bit 3 in the register */
197 /* Set the overflow flag to zero */
198 MXCSR = MXCSR & 0xFFF7;
211 #endif /* _gmx_x86_avx_128_fma_h_ */