/*
* Copyright 2000-2017 the Libav developers
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program 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 General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see .
*/
#include "config.h"
#include
#include "cpu.h"
#if ARCH_X86_64 && HAVE_NASM
static int 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;
mg2di_cpu_cpuid(0, &max_std_level, &vendor.i[0], &vendor.i[2], &vendor.i[1]);
if (max_std_level >= 1) {
mg2di_cpu_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 |= MG2DI_CPU_FLAG_CMOV;
if (std_caps & (1 << 23))
rval |= MG2DI_CPU_FLAG_MMX;
if (std_caps & (1 << 25))
rval |= MG2DI_CPU_FLAG_MMXEXT;
if (std_caps & (1 << 25))
rval |= MG2DI_CPU_FLAG_SSE;
if (std_caps & (1 << 26))
rval |= MG2DI_CPU_FLAG_SSE2;
if (ecx & 1)
rval |= MG2DI_CPU_FLAG_SSE3;
if (ecx & 0x00000200 )
rval |= MG2DI_CPU_FLAG_SSSE3;
if (ecx & 0x00080000 )
rval |= MG2DI_CPU_FLAG_SSE4;
if (ecx & 0x00100000 )
rval |= MG2DI_CPU_FLAG_SSE42;
/* Check OXSAVE and AVX bits */
if ((ecx & 0x18000000) == 0x18000000) {
/* Check for OS support */
mg2di_cpu_xgetbv(0, &eax, &edx);
if ((eax & 0x6) == 0x6) {
rval |= MG2DI_CPU_FLAG_AVX;
if (ecx & 0x00001000)
rval |= MG2DI_CPU_FLAG_FMA3;
}
}
}
if (max_std_level >= 7) {
mg2di_cpu_cpuid(7, &eax, &ebx, &ecx, &edx);
if (ebx & 0x00000020)
rval |= MG2DI_CPU_FLAG_AVX2;
/* BMI1/2 don't need OS support */
if (ebx & 0x00000008) {
rval |= MG2DI_CPU_FLAG_BMI1;
if (ebx & 0x00000100)
rval |= MG2DI_CPU_FLAG_BMI2;
}
}
mg2di_cpu_cpuid(0x80000000, &max_ext_level, &ebx, &ecx, &edx);
if (max_ext_level >= 0x80000001) {
mg2di_cpu_cpuid(0x80000001, &eax, &ebx, &ecx, &ext_caps);
if (ext_caps & (1U << 31))
rval |= MG2DI_CPU_FLAG_3DNOW;
if (ext_caps & (1 << 30))
rval |= MG2DI_CPU_FLAG_3DNOWEXT;
if (ext_caps & (1 << 23))
rval |= MG2DI_CPU_FLAG_MMX;
if (ext_caps & (1 << 22))
rval |= MG2DI_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.
MG2DI_CPU_FLAG_SSE2 and MG2DI_CPU_FLAG_SSE2SLOW are both set in this case
so that SSE2 is used unless explicitly disabled by checking
MG2DI_CPU_FLAG_SSE2SLOW. */
if (rval & MG2DI_CPU_FLAG_SSE2 && !(ecx & 0x00000040))
rval |= MG2DI_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
based CPUs as they lack 256-bit execution units. SSE/AVX functions
using XMM registers are always faster on them.
MG2DI_CPU_FLAG_AVX and MG2DI_CPU_FLAG_AVXSLOW are both set so that AVX is
used unless explicitly disabled by checking MG2DI_CPU_FLAG_AVXSLOW.
TODO: Confirm if Excavator is affected or not by this once it's
released, and update the check if necessary. Same for btver2. */
if (family == 0x15 && (rval & MG2DI_CPU_FLAG_AVX))
rval |= MG2DI_CPU_FLAG_AVXSLOW;
}
/* XOP and FMA4 use the AVX instruction coding scheme, so they can't be
* used unless the OS has AVX support. */
if (rval & MG2DI_CPU_FLAG_AVX) {
if (ecx & 0x00000800)
rval |= MG2DI_CPU_FLAG_XOP;
if (ecx & 0x00010000)
rval |= MG2DI_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.
* MG2DI_CPU_FLAG_SSE2 is disabled and MG2DI_CPU_FLAG_SSE2SLOW is
* enabled so that SSE2 is not used unless explicitly enabled
* by checking MG2DI_CPU_FLAG_SSE2SLOW. The same situation
* applies for MG2DI_CPU_FLAG_SSE3 and MG2DI_CPU_FLAG_SSE3SLOW. */
if (rval & MG2DI_CPU_FLAG_SSE2)
rval ^= MG2DI_CPU_FLAG_SSE2SLOW | MG2DI_CPU_FLAG_SSE2;
if (rval & MG2DI_CPU_FLAG_SSE3)
rval ^= MG2DI_CPU_FLAG_SSE3SLOW | MG2DI_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 |= MG2DI_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 & MG2DI_CPU_FLAG_SSSE3) && !(rval & MG2DI_CPU_FLAG_SSE4) &&
family == 6 && model < 23)
rval |= MG2DI_CPU_FLAG_SSSE3SLOW;
}
return rval;
}
#endif
int mg2di_cpu_flags_get(void)
{
int flags = 0;
#if ARCH_X86_64 && HAVE_NASM
flags = get_cpu_flags_x86();
#endif
return flags;
}