|
cannam@95
|
1 /*
|
|
cannam@95
|
2 * Copyright (c) 2003, 2007-11 Matteo Frigo
|
|
cannam@95
|
3 * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
|
|
cannam@95
|
4 *
|
|
cannam@95
|
5 * This program is free software; you can redistribute it and/or modify
|
|
cannam@95
|
6 * it under the terms of the GNU General Public License as published by
|
|
cannam@95
|
7 * the Free Software Foundation; either version 2 of the License, or
|
|
cannam@95
|
8 * (at your option) any later version.
|
|
cannam@95
|
9 *
|
|
cannam@95
|
10 * This program is distributed in the hope that it will be useful,
|
|
cannam@95
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
cannam@95
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
cannam@95
|
13 * GNU General Public License for more details.
|
|
cannam@95
|
14 *
|
|
cannam@95
|
15 * You should have received a copy of the GNU General Public License
|
|
cannam@95
|
16 * along with this program; if not, write to the Free Software
|
|
cannam@95
|
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
cannam@95
|
18 *
|
|
cannam@95
|
19 */
|
|
cannam@95
|
20
|
|
cannam@95
|
21
|
|
cannam@95
|
22 #include "ifftw.h"
|
|
cannam@95
|
23
|
|
cannam@95
|
24 #ifdef FFTW_SINGLE
|
|
cannam@95
|
25 # define DS(d,s) s /* single-precision option */
|
|
cannam@95
|
26 #else
|
|
cannam@95
|
27 # define DS(d,s) d /* double-precision option */
|
|
cannam@95
|
28 #endif
|
|
cannam@95
|
29
|
|
cannam@95
|
30 #if HAVE_SSE2
|
|
cannam@95
|
31
|
|
cannam@95
|
32 # if defined(__x86_64__) || defined(_M_X64) || defined(_M_AMD64)
|
|
cannam@95
|
33
|
|
cannam@95
|
34 int X(have_simd_sse2)(void)
|
|
cannam@95
|
35 {
|
|
cannam@95
|
36 return 1;
|
|
cannam@95
|
37 }
|
|
cannam@95
|
38
|
|
cannam@95
|
39 # else /* !x86_64 */
|
|
cannam@95
|
40
|
|
cannam@95
|
41 # include <signal.h>
|
|
cannam@95
|
42 # include <setjmp.h>
|
|
cannam@95
|
43 # include "x86-cpuid.h"
|
|
cannam@95
|
44
|
|
cannam@95
|
45 static jmp_buf jb;
|
|
cannam@95
|
46
|
|
cannam@95
|
47 static void sighandler(int x)
|
|
cannam@95
|
48 {
|
|
cannam@95
|
49 UNUSED(x);
|
|
cannam@95
|
50 longjmp(jb, 1);
|
|
cannam@95
|
51 }
|
|
cannam@95
|
52
|
|
cannam@95
|
53 static int sse2_works(void)
|
|
cannam@95
|
54 {
|
|
cannam@95
|
55 void (*oldsig)(int);
|
|
cannam@95
|
56 oldsig = signal(SIGILL, sighandler);
|
|
cannam@95
|
57 if (setjmp(jb)) {
|
|
cannam@95
|
58 signal(SIGILL, oldsig);
|
|
cannam@95
|
59 return 0;
|
|
cannam@95
|
60 } else {
|
|
cannam@95
|
61 # ifdef _MSC_VER
|
|
cannam@95
|
62 _asm { DS(xorpd,xorps) xmm0,xmm0 }
|
|
cannam@95
|
63 # else
|
|
cannam@95
|
64 /* asm volatile ("xorpd/s %xmm0, %xmm0"); */
|
|
cannam@95
|
65 asm volatile(DS(".byte 0x66; .byte 0x0f; .byte 0x57; .byte 0xc0",
|
|
cannam@95
|
66 ".byte 0x0f; .byte 0x57; .byte 0xc0"));
|
|
cannam@95
|
67 # endif
|
|
cannam@95
|
68 signal(SIGILL, oldsig);
|
|
cannam@95
|
69 return 1;
|
|
cannam@95
|
70 }
|
|
cannam@95
|
71 }
|
|
cannam@95
|
72
|
|
cannam@95
|
73 extern void X(check_alignment_of_sse2_pm)(void);
|
|
cannam@95
|
74
|
|
cannam@95
|
75 int X(have_simd_sse2)(void)
|
|
cannam@95
|
76 {
|
|
cannam@95
|
77 static int init = 0, res;
|
|
cannam@95
|
78
|
|
cannam@95
|
79 if (!init) {
|
|
cannam@95
|
80 res = !is_386()
|
|
cannam@95
|
81 && has_cpuid()
|
|
cannam@95
|
82 && (cpuid_edx(1) & (1 << DS(26,25)))
|
|
cannam@95
|
83 && sse2_works();
|
|
cannam@95
|
84 init = 1;
|
|
cannam@95
|
85 X(check_alignment_of_sse2_pm)();
|
|
cannam@95
|
86 }
|
|
cannam@95
|
87 return res;
|
|
cannam@95
|
88 }
|
|
cannam@95
|
89
|
|
cannam@95
|
90 # endif
|
|
cannam@95
|
91
|
|
cannam@95
|
92 #endif
|
