1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
|
/*
* CPU detection code, extracted from mmx.h
* (c)1997-99 by H. Dietz and R. Fisher
* Converted to C and improved by Fabrice Bellard.
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include <stdlib.h>
#include <string.h>
#include "libavutil/x86/asm.h"
#include "libavutil/cpu.h"
#if HAVE_INLINE_ASM
/* ebx saving is necessary for PIC. gcc seems unable to see it alone */
#define cpuid(index, eax, ebx, ecx, edx) \
__asm__ volatile ( \
"mov %%"REG_b", %%"REG_S" \n\t" \
"cpuid \n\t" \
"xchg %%"REG_b", %%"REG_S \
: "=a" (eax), "=S" (ebx), "=c" (ecx), "=d" (edx) \
: "0" (index))
#elif HAVE_CPUID
#include <intrin.h>
#define cpuid(index, eax, ebx, ecx, edx) \
do { \
int info[4]; \
__cpuid(info, index); \
eax = info[0]; \
ebx = info[1]; \
ecx = info[2]; \
edx = info[3]; \
} while (0)
#endif /* HAVE_CPUID */
#if HAVE_INLINE_ASM
#define xgetbv(index, eax, edx) \
__asm__ (".byte 0x0f, 0x01, 0xd0" : "=a"(eax), "=d"(edx) : "c" (index))
#elif HAVE_XGETBV
#include <immintrin.h>
#define xgetbv(index, eax, edx) \
do { \
uint64_t res = __xgetbv(index); \
eax = res; \
edx = res >> 32; \
} while (0)
#endif /* HAVE_XGETBV */
#if HAVE_INLINE_ASM
#define get_eflags(x) \
__asm__ volatile ("pushfl \n" \
"pop %0 \n" \
: "=r"(x))
#define set_eflags(x) \
__asm__ volatile ("push %0 \n" \
"popfl \n" \
:: "r"(x))
#elif HAVE_RWEFLAGS
#include <intrin.h>
#define get_eflags(x) \
x = __readeflags()
#define set_eflags(x) \
__writeeflags(x)
#endif /* HAVE_INLINE_ASM */
/* Function to test if multimedia instructions are supported... */
int ff_get_cpu_flags_x86(void)
{
int rval = 0;
int eax, ebx, ecx, edx;
int max_std_level, max_ext_level, std_caps = 0, ext_caps = 0;
int family = 0, model = 0;
union { int i[3]; char c[12]; } vendor;
#if ARCH_X86_32
x86_reg a, c;
/* Check if CPUID is supported by attempting to toggle the ID bit in
* the EFLAGS register. */
get_eflags(a);
set_eflags(a ^ 0x200000);
get_eflags(c);
if (a == c)
return 0; /* CPUID not supported */
#endif
cpuid(0, max_std_level, ebx, ecx, edx);
vendor.i[0] = ebx;
vendor.i[1] = edx;
vendor.i[2] = ecx;
if (max_std_level >= 1) {
cpuid(1, eax, ebx, ecx, std_caps);
family = ((eax >> 8) & 0xf) + ((eax >> 20) & 0xff);
model = ((eax >> 4) & 0xf) + ((eax >> 12) & 0xf0);
if (std_caps & (1 << 15))
rval |= AV_CPU_FLAG_CMOV;
if (std_caps & (1 << 23))
rval |= AV_CPU_FLAG_MMX;
if (std_caps & (1 << 25))
rval |= AV_CPU_FLAG_MMXEXT;
#if HAVE_SSE
if (std_caps & (1 << 25))
rval |= AV_CPU_FLAG_SSE;
if (std_caps & (1 << 26))
rval |= AV_CPU_FLAG_SSE2;
if (ecx & 1)
rval |= AV_CPU_FLAG_SSE3;
if (ecx & 0x00000200 )
rval |= AV_CPU_FLAG_SSSE3;
if (ecx & 0x00080000 )
rval |= AV_CPU_FLAG_SSE4;
if (ecx & 0x00100000 )
rval |= AV_CPU_FLAG_SSE42;
#if HAVE_AVX
/* Check OXSAVE and AVX bits */
if ((ecx & 0x18000000) == 0x18000000) {
/* Check for OS support */
xgetbv(0, eax, edx);
if ((eax & 0x6) == 0x6)
rval |= AV_CPU_FLAG_AVX;
}
#endif /* HAVE_AVX */
#endif /* HAVE_SSE */
}
cpuid(0x80000000, max_ext_level, ebx, ecx, edx);
if (max_ext_level >= 0x80000001) {
cpuid(0x80000001, eax, ebx, ecx, ext_caps);
if (ext_caps & (1U << 31))
rval |= AV_CPU_FLAG_3DNOW;
if (ext_caps & (1 << 30))
rval |= AV_CPU_FLAG_3DNOWEXT;
if (ext_caps & (1 << 23))
rval |= AV_CPU_FLAG_MMX;
if (ext_caps & (1 << 22))
rval |= AV_CPU_FLAG_MMXEXT;
/* Allow for selectively disabling SSE2 functions on AMD processors
with SSE2 support but not SSE4a. This includes Athlon64, some
Opteron, and some Sempron processors. MMX, SSE, or 3DNow! are faster
than SSE2 often enough to utilize this special-case flag.
AV_CPU_FLAG_SSE2 and AV_CPU_FLAG_SSE2SLOW are both set in this case
so that SSE2 is used unless explicitly disabled by checking
AV_CPU_FLAG_SSE2SLOW. */
if (!strncmp(vendor.c, "AuthenticAMD", 12) &&
rval & AV_CPU_FLAG_SSE2 && !(ecx & 0x00000040)) {
rval |= AV_CPU_FLAG_SSE2SLOW;
}
/* XOP and FMA4 use the AVX instruction coding scheme, so they can't be
* used unless the OS has AVX support. */
if (rval & AV_CPU_FLAG_AVX) {
if (ecx & 0x00000800)
rval |= AV_CPU_FLAG_XOP;
if (ecx & 0x00010000)
rval |= AV_CPU_FLAG_FMA4;
}
}
if (!strncmp(vendor.c, "GenuineIntel", 12)) {
if (family == 6 && (model == 9 || model == 13 || model == 14)) {
/* 6/9 (pentium-m "banias"), 6/13 (pentium-m "dothan"), and
* 6/14 (core1 "yonah") theoretically support sse2, but it's
* usually slower than mmx, so let's just pretend they don't.
* AV_CPU_FLAG_SSE2 is disabled and AV_CPU_FLAG_SSE2SLOW is
* enabled so that SSE2 is not used unless explicitly enabled
* by checking AV_CPU_FLAG_SSE2SLOW. The same situation
* applies for AV_CPU_FLAG_SSE3 and AV_CPU_FLAG_SSE3SLOW. */
if (rval & AV_CPU_FLAG_SSE2)
rval ^= AV_CPU_FLAG_SSE2SLOW | AV_CPU_FLAG_SSE2;
if (rval & AV_CPU_FLAG_SSE3)
rval ^= AV_CPU_FLAG_SSE3SLOW | AV_CPU_FLAG_SSE3;
}
/* The Atom processor has SSSE3 support, which is useful in many cases,
* but sometimes the SSSE3 version is slower than the SSE2 equivalent
* on the Atom, but is generally faster on other processors supporting
* SSSE3. This flag allows for selectively disabling certain SSSE3
* functions on the Atom. */
if (family == 6 && model == 28)
rval |= AV_CPU_FLAG_ATOM;
}
return rval;
}
|
