eden-emu
/
eden
mirror of https://git.eden-emu.dev/eden-emu/eden.git
1
0
Fork 0
Code Releases Wiki Activity
Browse Source

[dynarmic] fix perf regression due to hard CPUID queries (#4328) 

Signed-off-by: lizzie <lizzie@eden-emu.dev>

- [x] I have read and followed the [Contribution Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/CONTRIBUTING.md#code-contributions).
- [x] I have read and followed the [AI Policy](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/AI.md)
- [x] I have read and followed the [Coding Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/Coding.md) to the best of my ability.

-------------------

should fix whatever issue bruno had on Windows

basically CPUID ctor object for xbyak is very VERY heavy, so we just move that to a static we do once on program startup instead!

Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4328
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>
Reviewed-by: Maufeat <sahyno1996@gmail.com>
xbzk/android-enable-extended-logging
lizzie 1 day ago
committed by crueter
parent
f7e5ceeacb
commit
cb9ec1dc1a
No known key found for this signature in database GPG Key ID: 425ACD2D4830EBC6
1 changed files with 30 additions and 50 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 80
      src/dynarmic/src/dynarmic/backend/x64/block_of_code.cpp

80
src/dynarmic/src/dynarmic/backend/x64/block_of_code.cpp
View File

@ -76,57 +76,34 @@ void ProtectMemory(const void* base, size_t size, bool is_executable) {
}
#endif
HostFeature GetHostFeatures() {
HostFeature features = {};
static const HostFeature features = []() {
HostFeature f{};
#ifdef DYNARMIC_ENABLE_CPU_FEATURE_DETECTION
using Cpu = Xbyak::util::Cpu;
Xbyak::util::Cpu cpu_info{};
if (cpu_info.has(Cpu::tSSSE3))
features |= HostFeature::SSSE3;
if (cpu_info.has(Cpu::tSSE41))
features |= HostFeature::SSE41;
if (cpu_info.has(Cpu::tSSE42))
features |= HostFeature::SSE42;
if (cpu_info.has(Cpu::tAVX))
features |= HostFeature::AVX;
if (cpu_info.has(Cpu::tAVX2))
features |= HostFeature::AVX2;
if (cpu_info.has(Cpu::tAVX512F))
features |= HostFeature::AVX512F;
if (cpu_info.has(Cpu::tAVX512CD))
features |= HostFeature::AVX512CD;
if (cpu_info.has(Cpu::tAVX512VL))
features |= HostFeature::AVX512VL;
if (cpu_info.has(Cpu::tAVX512BW))
features |= HostFeature::AVX512BW;
if (cpu_info.has(Cpu::tAVX512DQ))
features |= HostFeature::AVX512DQ;
if (cpu_info.has(Cpu::tAVX512_BITALG))
features |= HostFeature::AVX512BITALG;
if (cpu_info.has(Cpu::tAVX512VBMI))
features |= HostFeature::AVX512VBMI;
if (cpu_info.has(Cpu::tPCLMULQDQ))
features |= HostFeature::PCLMULQDQ;
if (cpu_info.has(Cpu::tF16C))
features |= HostFeature::F16C;
if (cpu_info.has(Cpu::tFMA))
features |= HostFeature::FMA;
if (cpu_info.has(Cpu::tAESNI))
features |= HostFeature::AES;
if (cpu_info.has(Cpu::tSHA))
features |= HostFeature::SHA;
if (cpu_info.has(Cpu::tPOPCNT))
features |= HostFeature::POPCNT;
if (cpu_info.has(Cpu::tBMI1))
features |= HostFeature::BMI1;
if (cpu_info.has(Cpu::tBMI2))
features |= HostFeature::BMI2;
if (cpu_info.has(Cpu::tLZCNT))
features |= HostFeature::LZCNT;
if (cpu_info.has(Cpu::tGFNI))
features |= HostFeature::GFNI;
if (cpu_info.has(Cpu::tWAITPKG))
features |= HostFeature::WAITPKG;
if (cpu_info.has(Cpu::tSSSE3)) f |= HostFeature::SSSE3;
if (cpu_info.has(Cpu::tSSE41)) f |= HostFeature::SSE41;
if (cpu_info.has(Cpu::tSSE42)) f |= HostFeature::SSE42;
if (cpu_info.has(Cpu::tAVX)) f |= HostFeature::AVX;
if (cpu_info.has(Cpu::tAVX2)) f |= HostFeature::AVX2;
if (cpu_info.has(Cpu::tAVX512F)) f |= HostFeature::AVX512F;
if (cpu_info.has(Cpu::tAVX512CD)) f |= HostFeature::AVX512CD;
if (cpu_info.has(Cpu::tAVX512VL)) f |= HostFeature::AVX512VL;
if (cpu_info.has(Cpu::tAVX512BW)) f |= HostFeature::AVX512BW;
if (cpu_info.has(Cpu::tAVX512DQ)) f |= HostFeature::AVX512DQ;
if (cpu_info.has(Cpu::tAVX512_BITALG)) f |= HostFeature::AVX512BITALG;
if (cpu_info.has(Cpu::tAVX512VBMI)) f |= HostFeature::AVX512VBMI;
if (cpu_info.has(Cpu::tPCLMULQDQ)) f |= HostFeature::PCLMULQDQ;
if (cpu_info.has(Cpu::tF16C)) f |= HostFeature::F16C;
if (cpu_info.has(Cpu::tFMA)) f |= HostFeature::FMA;
if (cpu_info.has(Cpu::tAESNI)) f |= HostFeature::AES;
if (cpu_info.has(Cpu::tSHA)) f |= HostFeature::SHA;
if (cpu_info.has(Cpu::tPOPCNT)) f |= HostFeature::POPCNT;
if (cpu_info.has(Cpu::tBMI1)) f |= HostFeature::BMI1;
if (cpu_info.has(Cpu::tBMI2)) f |= HostFeature::BMI2;
if (cpu_info.has(Cpu::tLZCNT)) f |= HostFeature::LZCNT;
if (cpu_info.has(Cpu::tGFNI)) f |= HostFeature::GFNI;
if (cpu_info.has(Cpu::tWAITPKG)) f |= HostFeature::WAITPKG;
if (cpu_info.has(Cpu::tBMI2)) {
// BMI2 instructions such as pdep and pext have been very slow up until Zen 3.
// Check for Zen 3 or newer by its family (0x19).
@ -138,12 +115,15 @@ HostFeature GetHostFeatures() {
const u32 family_extended = mcl::bit::get_bits<20, 27>(data[0]);
const u32 family = family_base + family_extended;
if (family >= 0x19)
features |= HostFeature::FastBMI2;
f |= HostFeature::FastBMI2;
} else {
features |= HostFeature::FastBMI2;
f |= HostFeature::FastBMI2;
}
}
return f;
#endif
}();
HostFeature GetHostFeatures() {
return features;
}

Write Preview
Loading…
Cancel
Save