Compare commits

..

8 Commits

Author SHA1 Message Date
lizzie f8ccdb00f2 fix 2026-08-31 03:42:51 +00:00
lizzie 050af0bc51 Revert "[dynarmic] fix windows perf regression due to reordered CPR"
This reverts commit 403cc0ab66.
2026-08-31 03:35:08 +00:00
lizzie e322e34148 Revert "goto older xbyak"
This reverts commit 2e9d97aa82.
2026-08-31 03:35:01 +00:00
lizzie 804c1dcfaf Revert "stateless allocator?"
This reverts commit d918efbea5.
2026-08-31 03:34:51 +00:00
lizzie 5edb9d37ef cpu features stored in static oncemore 2026-08-31 03:34:46 +00:00
lizzie d918efbea5 stateless allocator? 2026-08-31 03:27:12 +00:00
lizzie 2e9d97aa82 goto older xbyak 2026-08-31 03:01:39 +00:00
lizzie 403cc0ab66 [dynarmic] fix windows perf regression due to reordered CPR
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-08-31 02:42:30 +00:00
4 changed files with 39 additions and 58 deletions
+2 -1
View File
@@ -25,13 +25,14 @@
#include "common/container/unordered_map.h"
#include "common/container/unordered_set.h"
#include <boost/unordered_map.hpp>
#define XBYAK_NO_EXCEPTION 1
#define XBYAK_STD_UNORDERED_SET ::Common::unordered_set
#define XBYAK_STD_UNORDERED_MAP ::Common::unordered_map
#define XBYAK_STD_UNORDERED_MULTIMAP boost::unordered_multimap
#include <xbyak/xbyak.h>
#include <xbyak/xbyak_util.h>
#include <xbyak/xbyak.h>
namespace Common::X64 {
constexpr size_t RegToIndex(const Xbyak::Reg& reg) {
+1 -5
View File
@@ -95,6 +95,7 @@ else()
-pedantic
-Wno-missing-braces
-fno-rtti
#-fno-exceptions
)
if (CXX_GCC)
# GCC produces bogus -Warray-bounds warnings from xbyak headers for code paths that are not
@@ -112,11 +113,6 @@ else()
# Clang mistakenly blames CMake for using unused arguments during compilation
list(APPEND DYNARMIC_CXX_FLAGS -Wno-unused-command-line-argument)
endif()
# TODO: oaknut exceptions
# TODO: fix chromeOS
if ("x86_64" IN_LIST ARCHITECTURE AND NOT ANDROID)
list(APPEND DYNARMIC_CXX_FLAGS -fno-exceptions)
endif()
endif()
if (NOT Boost_FOUND)
@@ -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;
}
@@ -465,8 +445,13 @@ size_t BlockOfCode::GetTotalCodeSize() const {
}
void* BlockOfCode::AllocateFromCodeSpace(size_t alloc_size) {
ASSERT(size_ + alloc_size < maxSize_);
if (size_ + alloc_size >= maxSize_) {
using Xbyak::Error;
XBYAK_THROW(Xbyak::ERR_CODE_IS_TOO_BIG);
}
EnsureMemoryCommitted(alloc_size);
void* ret = getCurr<void*>();
size_ += alloc_size;
memset(ret, 0, alloc_size);
@@ -7,7 +7,6 @@
#include "common/container/unordered_map.h"
#include "common/container/unordered_set.h"
#include <boost/unordered_map.hpp>
#define XBYAK_NO_EXCEPTION 1
#define XBYAK_STD_UNORDERED_SET ::Common::unordered_set
#define XBYAK_STD_UNORDERED_MAP ::Common::unordered_map
#define XBYAK_STD_UNORDERED_MULTIMAP boost::unordered_multimap