Chris@42: /*
Chris@42:  * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@42:  * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@42:  *
Chris@42:  * This program is free software; you can redistribute it and/or modify
Chris@42:  * it under the terms of the GNU General Public License as published by
Chris@42:  * the Free Software Foundation; either version 2 of the License, or
Chris@42:  * (at your option) any later version.
Chris@42:  *
Chris@42:  * This program is distributed in the hope that it will be useful,
Chris@42:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@42:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@42:  * GNU General Public License for more details.
Chris@42:  *
Chris@42:  * You should have received a copy of the GNU General Public License
Chris@42:  * along with this program; if not, write to the Free Software
Chris@42:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@42:  *
Chris@42:  */
Chris@42: 
Chris@42: 
Chris@42: #include "ifftw.h"
Chris@42: 
Chris@42: #if HAVE_AVX2
Chris@42: 
Chris@42: #if defined(__x86_64__) || defined(_M_X64) || defined(_M_AMD64)
Chris@42: #    include "amd64-cpuid.h"
Chris@42: #else
Chris@42: #    include "x86-cpuid.h"
Chris@42: #endif
Chris@42: 
Chris@42: int X(have_simd_avx2_128)(void)
Chris@42: {
Chris@42:     static int init = 0, res;
Chris@42:     int max_stdfn, eax, ebx, ecx, edx;
Chris@42:     
Chris@42:     if (!init) {
Chris@42:          cpuid_all(0,0,&eax,&ebx,&ecx,&edx);
Chris@42:          max_stdfn = eax;
Chris@42:          if (max_stdfn >= 0x1) {
Chris@42:               /* have AVX and OSXSAVE? (implies XGETBV exists) */
Chris@42:               cpuid_all(0x1, 0, &eax, &ebx, &ecx, &edx);
Chris@42:               if ((ecx & 0x18000000) == 0x18000000) {              
Chris@42:                    /* have AVX2? */
Chris@42:                    cpuid_all(7,0,&eax,&ebx,&ecx,&edx);
Chris@42:                    if (ebx & (1 << 5)) {
Chris@42:                         /* have OS support for XMM, YMM? */
Chris@42:                         res = ((xgetbv_eax(0) & 0x6) == 0x6);
Chris@42:                    }
Chris@42:               }
Chris@42:         }
Chris@42:         init = 1;
Chris@42:     }
Chris@42:     return res;
Chris@42: }
Chris@42: 
Chris@42: int X(have_simd_avx2)(void)
Chris@42: {
Chris@42:   /*
Chris@42:    * For now 256-bit AVX2 support is identical to 128-bit.
Chris@42:    * This might change in the future if AMD released AVX2-capable
Chris@42:    * chips that work better with the 128-bit flavor, but since AMD
Chris@42:    * might actually change it to implement 256-bit AVX2 efficiently
Chris@42:    * by then we don't want to disable it before we know.
Chris@42:    */
Chris@42:   return X(have_simd_avx2_128)();
Chris@42: }
Chris@42: #endif