ffmpeg/libavutil/x86/cpu.c

285 lines
10 KiB
C

/*
* 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/x86/cpu.h"
#include "libavutil/cpu.h"
#include "libavutil/cpu_internal.h"
#if HAVE_X86ASM
#define cpuid(index, eax, ebx, ecx, edx) \
ff_cpu_cpuid(index, &eax, &ebx, &ecx, &edx)
#define xgetbv(index, eax, edx) \
ff_cpu_xgetbv(index, &eax, &edx)
#elif 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 %%"FF_REG_b", %%"FF_REG_S" \n\t" \
"cpuid \n\t" \
"xchg %%"FF_REG_b", %%"FF_REG_S \
: "=a" (eax), "=S" (ebx), "=c" (ecx), "=d" (edx) \
: "0" (index), "2"(0))
#define xgetbv(index, eax, edx) \
__asm__ (".byte 0x0f, 0x01, 0xd0" : "=a"(eax), "=d"(edx) : "c" (index))
#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))
#endif /* HAVE_INLINE_ASM */
#if ARCH_X86_64
#define cpuid_test() 1
#elif HAVE_X86ASM
#define cpuid_test ff_cpu_cpuid_test
#elif HAVE_INLINE_ASM
static int cpuid_test(void)
{
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);
return a != c;
}
#endif
/* Function to test if multimedia instructions are supported... */
int ff_get_cpu_flags_x86(void)
{
int rval = 0;
#ifdef cpuid
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;
int xcr0_lo = 0, xcr0_hi = 0;
if (!cpuid_test())
return 0; /* CPUID not supported */
cpuid(0, max_std_level, vendor.i[0], vendor.i[2], vendor.i[1]);
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 (ecx & 0x02000000 )
rval |= AV_CPU_FLAG_AESNI;
#if HAVE_AVX
/* Check OXSAVE and AVX bits */
if ((ecx & 0x18000000) == 0x18000000) {
/* Check for OS support */
xgetbv(0, xcr0_lo, xcr0_hi);
if ((xcr0_lo & 0x6) == 0x6) {
rval |= AV_CPU_FLAG_AVX;
if (ecx & 0x00001000)
rval |= AV_CPU_FLAG_FMA3;
}
}
#endif /* HAVE_AVX */
#endif /* HAVE_SSE */
}
if (max_std_level >= 7) {
cpuid(7, eax, ebx, ecx, edx);
#if HAVE_AVX2
if ((rval & AV_CPU_FLAG_AVX) && (ebx & 0x00000020))
rval |= AV_CPU_FLAG_AVX2;
#if HAVE_AVX512 /* F, CD, BW, DQ, VL */
if ((xcr0_lo & 0xe0) == 0xe0) { /* OPMASK/ZMM state */
if ((rval & AV_CPU_FLAG_AVX2) && (ebx & 0xd0030000) == 0xd0030000) {
rval |= AV_CPU_FLAG_AVX512;
#if HAVE_AVX512ICL
if ((ebx & 0xd0200000) == 0xd0200000 && (ecx & 0x5f42) == 0x5f42)
rval |= AV_CPU_FLAG_AVX512ICL;
#endif /* HAVE_AVX512ICL */
}
}
#endif /* HAVE_AVX512 */
#endif /* HAVE_AVX2 */
/* BMI1/2 don't need OS support */
if (ebx & 0x00000008) {
rval |= AV_CPU_FLAG_BMI1;
if (ebx & 0x00000100)
rval |= AV_CPU_FLAG_BMI2;
}
}
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;
if (!strncmp(vendor.c, "AuthenticAMD", 12)) {
/* 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 (rval & AV_CPU_FLAG_SSE2 && !(ecx & 0x00000040))
rval |= AV_CPU_FLAG_SSE2SLOW;
/* Similar to the above but for AVX functions on AMD processors.
This is necessary only for functions using YMM registers on Bulldozer
and Jaguar based CPUs as they lack 256-bit execution units. SSE/AVX
functions using XMM registers are always faster on them.
AV_CPU_FLAG_AVX and AV_CPU_FLAG_AVXSLOW are both set so that AVX is
used unless explicitly disabled by checking AV_CPU_FLAG_AVXSLOW. */
if ((family == 0x15 || family == 0x16) && (rval & AV_CPU_FLAG_AVX))
rval |= AV_CPU_FLAG_AVXSLOW;
/* Zen 3 and earlier have slow gather */
if ((family <= 0x19) && (rval & AV_CPU_FLAG_AVX2))
rval |= AV_CPU_FLAG_SLOW_GATHER;
}
/* 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;
/* Conroe has a slow shuffle unit. Check the model number to ensure not
* to include crippled low-end Penryns and Nehalems that lack SSE4. */
if ((rval & AV_CPU_FLAG_SSSE3) && !(rval & AV_CPU_FLAG_SSE4) &&
family == 6 && model < 23)
rval |= AV_CPU_FLAG_SSSE3SLOW;
/* Haswell has slow gather */
if ((rval & AV_CPU_FLAG_AVX2) && family == 6 && model < 70)
rval |= AV_CPU_FLAG_SLOW_GATHER;
}
#endif /* cpuid */
return rval;
}
size_t ff_get_cpu_max_align_x86(void)
{
int flags = av_get_cpu_flags();
if (flags & AV_CPU_FLAG_AVX512)
return 64;
if (flags & (AV_CPU_FLAG_AVX2 |
AV_CPU_FLAG_AVX |
AV_CPU_FLAG_XOP |
AV_CPU_FLAG_FMA4 |
AV_CPU_FLAG_FMA3 |
AV_CPU_FLAG_AVXSLOW))
return 32;
if (flags & (AV_CPU_FLAG_AESNI |
AV_CPU_FLAG_SSE42 |
AV_CPU_FLAG_SSE4 |
AV_CPU_FLAG_SSSE3 |
AV_CPU_FLAG_SSE3 |
AV_CPU_FLAG_SSE2 |
AV_CPU_FLAG_SSE |
AV_CPU_FLAG_ATOM |
AV_CPU_FLAG_SSSE3SLOW |
AV_CPU_FLAG_SSE3SLOW |
AV_CPU_FLAG_SSE2SLOW))
return 16;
return 8;
}