dolphin/Source/Core/Common/x64CPUDetect.cpp
Fiora 72c96c20d3 JIT: more optimizing of float ops based on known input characteristics
If the inputs are both float singles, and the top half is known to be identical
to the bottom half, we can use packed arithmetic instead of scalar to skip
the movddup.

This is slower on a few rather old CPUs, plus the Atom+Silvermont, so detect
Atom and disable it in that case.

Also avoid PPC_FP on stores if we know that the output came from a float op.
2014-11-29 11:33:11 -08:00

265 lines
6.3 KiB
C++

// Copyright 2013 Dolphin Emulator Project
// Licensed under GPLv2
// Refer to the license.txt file included.
#include <cstring>
#include <string>
#include "Common/CommonTypes.h"
#include "Common/CPUDetect.h"
#ifdef _WIN32
#include <intrin.h>
#else
//#include <config/i386/cpuid.h>
#include <xmmintrin.h>
#if defined __FreeBSD__
#include <sys/types.h>
#include <machine/cpufunc.h>
#else
static inline void do_cpuid(unsigned int *eax, unsigned int *ebx,
unsigned int *ecx, unsigned int *edx)
{
#if defined _LP64
// Note: EBX is reserved on Mac OS X and in PIC on Linux, so it has to
// restored at the end of the asm block.
__asm__ (
"cpuid;"
"movl %%ebx,%1;"
: "=a" (*eax),
"=S" (*ebx),
"=c" (*ecx),
"=d" (*edx)
: "a" (*eax)
: "rbx"
);
#else
__asm__ (
"cpuid;"
"movl %%ebx,%1;"
: "=a" (*eax),
"=S" (*ebx),
"=c" (*ecx),
"=d" (*edx)
: "a" (*eax)
: "ebx"
);
#endif
}
#endif /* defined __FreeBSD__ */
static void __cpuid(int info[4], int x)
{
#if defined __FreeBSD__
do_cpuid((unsigned int)x, (unsigned int*)info);
#else
unsigned int eax = x, ebx = 0, ecx = 0, edx = 0;
do_cpuid(&eax, &ebx, &ecx, &edx);
info[0] = eax;
info[1] = ebx;
info[2] = ecx;
info[3] = edx;
#endif
}
#define _XCR_XFEATURE_ENABLED_MASK 0
static unsigned long long _xgetbv(unsigned int index)
{
unsigned int eax, edx;
__asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
return ((unsigned long long)edx << 32) | eax;
}
#endif
CPUInfo cpu_info;
CPUInfo::CPUInfo()
{
Detect();
}
// Detects the various CPU features
void CPUInfo::Detect()
{
memset(this, 0, sizeof(*this));
#ifdef _M_X86_64
Mode64bit = true;
OS64bit = true;
#endif
num_cores = 1;
// Set obvious defaults, for extra safety
if (Mode64bit)
{
bSSE = true;
bSSE2 = true;
bLongMode = true;
}
// Assume CPU supports the CPUID instruction. Those that don't can barely
// boot modern OS:es anyway.
int cpu_id[4];
memset(brand_string, 0, sizeof(brand_string));
// Detect CPU's CPUID capabilities, and grab CPU string
__cpuid(cpu_id, 0x00000000);
u32 max_std_fn = cpu_id[0]; // EAX
*((int *)brand_string) = cpu_id[1];
*((int *)(brand_string + 4)) = cpu_id[3];
*((int *)(brand_string + 8)) = cpu_id[2];
__cpuid(cpu_id, 0x80000000);
u32 max_ex_fn = cpu_id[0];
if (!strcmp(brand_string, "GenuineIntel"))
vendor = VENDOR_INTEL;
else if (!strcmp(brand_string, "AuthenticAMD"))
vendor = VENDOR_AMD;
else
vendor = VENDOR_OTHER;
// Set reasonable default brand string even if brand string not available.
strcpy(cpu_string, brand_string);
// Detect family and other misc stuff.
bool ht = false;
HTT = ht;
logical_cpu_count = 1;
if (max_std_fn >= 1)
{
__cpuid(cpu_id, 0x00000001);
int family = ((cpu_id[0] >> 8) & 0xf) + ((cpu_id[0] >> 20) & 0xff);
int model = ((cpu_id[0] >> 4) & 0xf) + ((cpu_id[0] >> 12) & 0xf0);
// Detect people unfortunate enough to be running Dolphin on an Atom
if (family == 6 && (model == 0x1C || model == 0x26 ||model == 0x27 || model == 0x35 || model == 0x36 ||
model == 0x37 || model == 0x4A || model == 0x4D || model == 0x5A || model == 0x5D))
bAtom = true;
logical_cpu_count = (cpu_id[1] >> 16) & 0xFF;
ht = (cpu_id[3] >> 28) & 1;
if ((cpu_id[3] >> 25) & 1) bSSE = true;
if ((cpu_id[3] >> 26) & 1) bSSE2 = true;
if ((cpu_id[2]) & 1) bSSE3 = true;
if ((cpu_id[2] >> 9) & 1) bSSSE3 = true;
if ((cpu_id[2] >> 19) & 1) bSSE4_1 = true;
if ((cpu_id[2] >> 20) & 1) bSSE4_2 = true;
if ((cpu_id[2] >> 22) & 1) bMOVBE = true;
if ((cpu_id[2] >> 25) & 1) bAES = true;
if ((cpu_id[3] >> 24) & 1)
{
// We can use FXSAVE.
bFXSR = true;
}
// AVX support requires 3 separate checks:
// - Is the AVX bit set in CPUID?
// - Is the XSAVE bit set in CPUID?
// - XGETBV result has the XCR bit set.
if (((cpu_id[2] >> 28) & 1) && ((cpu_id[2] >> 27) & 1))
{
if ((_xgetbv(_XCR_XFEATURE_ENABLED_MASK) & 0x6) == 0x6)
{
bAVX = true;
if ((cpu_id[2] >> 12) & 1)
bFMA = true;
}
}
if (max_std_fn >= 7)
{
__cpuid(cpu_id, 0x00000007);
// careful; we can't enable AVX2 unless the XSAVE/XGETBV checks above passed
if ((cpu_id[1] >> 5) & 1)
bAVX2 = bAVX;
if ((cpu_id[1] >> 3) & 1)
bBMI1 = true;
if ((cpu_id[1] >> 8) & 1)
bBMI2 = true;
}
}
bFlushToZero = bSSE;
if (max_ex_fn >= 0x80000004)
{
// Extract CPU model string
__cpuid(cpu_id, 0x80000002);
memcpy(cpu_string, cpu_id, sizeof(cpu_id));
__cpuid(cpu_id, 0x80000003);
memcpy(cpu_string + 16, cpu_id, sizeof(cpu_id));
__cpuid(cpu_id, 0x80000004);
memcpy(cpu_string + 32, cpu_id, sizeof(cpu_id));
}
if (max_ex_fn >= 0x80000001)
{
// Check for more features.
__cpuid(cpu_id, 0x80000001);
if (cpu_id[2] & 1) bLAHFSAHF64 = true;
if ((cpu_id[2] >> 5) & 1) bLZCNT = true;
if ((cpu_id[3] >> 29) & 1) bLongMode = true;
}
num_cores = (logical_cpu_count == 0) ? 1 : logical_cpu_count;
if (max_ex_fn >= 0x80000008)
{
// Get number of cores. This is a bit complicated. Following AMD manual here.
__cpuid(cpu_id, 0x80000008);
int apic_id_core_id_size = (cpu_id[2] >> 12) & 0xF;
if (apic_id_core_id_size == 0)
{
if (ht)
{
// New mechanism for modern Intel CPUs.
if (vendor == VENDOR_INTEL)
{
__cpuid(cpu_id, 0x00000004);
int cores_x_package = ((cpu_id[0] >> 26) & 0x3F) + 1;
HTT = (cores_x_package < logical_cpu_count);
cores_x_package = ((logical_cpu_count % cores_x_package) == 0) ? cores_x_package : 1;
num_cores = (cores_x_package > 1) ? cores_x_package : num_cores;
logical_cpu_count /= cores_x_package;
}
}
} else
{
// Use AMD's new method.
num_cores = (cpu_id[2] & 0xFF) + 1;
}
}
}
// Turn the CPU info into a string we can show
std::string CPUInfo::Summarize()
{
std::string sum(cpu_string);
sum += " (";
sum += brand_string;
sum += ")";
if (bSSE) sum += ", SSE";
if (bSSE2)
{
sum += ", SSE2";
if (!bFlushToZero)
sum += " (but not DAZ!)";
}
if (bSSE3) sum += ", SSE3";
if (bSSSE3) sum += ", SSSE3";
if (bSSE4_1) sum += ", SSE4.1";
if (bSSE4_2) sum += ", SSE4.2";
if (HTT) sum += ", HTT";
if (bAVX) sum += ", AVX";
if (bAVX2) sum += ", AVX2";
if (bBMI1) sum += ", BMI1";
if (bBMI2) sum += ", BMI2";
if (bFMA) sum += ", FMA";
if (bAES) sum += ", AES";
if (bMOVBE) sum += ", MOVBE";
if (bLongMode) sum += ", 64-bit support";
return sum;
}