Plexi09/eden-miror
1
0
Fork 0
mirror of https://git.eden-emu.dev/eden-emu/eden synced 2026-05-17 01:37:01 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

[dynarmic, macroHLE] Use faster ankerl for xbyak maps (#3716)

Browse source 
the nominal std::unordered_map<> isn't enough to warrant it's continued usage in xbyak internal structures, thus using ankerl should greatly remove a lot of indirection/stdc++ specific overhead from the usually poorly performant std::unordered_map

Both dynarmic and macroHLE should benefit greatly from a less-stupid unordered_dense

This should speedup both CPU and shader compilation latency (NOT BY A GREAT MARGIN) just enough to make loading zones in ToTK less horrific

Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3716
Reviewed-by: crueter <crueter@eden-emu.dev>
This commit is contained in:
lizzie 2026-05-15 22:07:45 +02:00 • committed by crueter
parent 413c7543ba
commit 2f0f8a979c
No known key found for this signature in database
GPG key ID: 425ACD2D4830EBC6
11 changed files with 93 additions and 95 deletions
Download patch file Download diff file Expand all files Collapse all files

3
src/common/CMakeLists.txt
View file

@ -186,8 +186,7 @@ if(ARCHITECTURE_x86_64)
x64/cpu_wait.h
x64/rdtsc.cpp
x64/rdtsc.h
x64/xbyak_abi.h
x64/xbyak_util.h)
x64/xbyak.h)
target_link_libraries(common PRIVATE xbyak::xbyak)
endif()

72
src/common/x64/xbyak_abi.h → src/common/x64/xbyak.h
View file

@ -1,13 +1,37 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: 2016 Citra Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <type_traits>
#include <bitset>
#include <initializer_list>
#include <xbyak/xbyak.h>
#include "common/assert.h"
// xbyak hates human beings
#ifdef __GNUC__
#pragma GCC diagnostic ignored "-Wconversion"
#pragma GCC diagnostic ignored "-Wshadow"
#endif
#ifdef __clang__
#pragma clang diagnostic ignored "-Wconversion"
#pragma clang diagnostic ignored "-Wshadow"
#endif
// You must ensure this matches with src/common/x64/xbyak.h on root dir
#include <ankerl/unordered_dense.h>
#include <boost/unordered_map.hpp>
#define XBYAK_STD_UNORDERED_SET ankerl::unordered_dense::set
#define XBYAK_STD_UNORDERED_MAP ankerl::unordered_dense::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) {
@ -174,12 +198,13 @@ inline ABIFrameInfo ABI_CalculateFrameSize(std::bitset<32> regs, size_t rsp_alig
rsp_alignment -= subtraction;
subtraction += rsp_alignment & 0xF;
return ABIFrameInfo{static_cast<s32>(subtraction),
static_cast<s32>(subtraction - xmm_base_subtraction)};
return ABIFrameInfo{
s32(subtraction),
s32(subtraction - xmm_base_subtraction)
};
}
inline size_t ABI_PushRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
size_t rsp_alignment, size_t needed_frame_size = 0) {
inline size_t ABI_PushRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs, size_t rsp_alignment, size_t needed_frame_size = 0) {
auto frame_info = ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size);
for (size_t i = 0; i < regs.size(); ++i) {
@ -202,8 +227,7 @@ inline size_t ABI_PushRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::b
return ABI_SHADOW_SPACE;
}
inline void ABI_PopRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs,
size_t rsp_alignment, size_t needed_frame_size = 0) {
inline void ABI_PopRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bitset<32> regs, size_t rsp_alignment, size_t needed_frame_size = 0) {
auto frame_info = ABI_CalculateFrameSize(regs, rsp_alignment, needed_frame_size);
for (size_t i = 0; i < regs.size(); ++i) {
@ -226,4 +250,38 @@ inline void ABI_PopRegistersAndAdjustStack(Xbyak::CodeGenerator& code, std::bits
}
}
// Constants for use with cmpps/cmpss
enum {
CMP_EQ = 0,
CMP_LT = 1,
CMP_LE = 2,
CMP_UNORD = 3,
CMP_NEQ = 4,
CMP_NLT = 5,
CMP_NLE = 6,
CMP_ORD = 7,
};
constexpr bool IsWithin2G(uintptr_t ref, uintptr_t target) {
const u64 distance = target - (ref + 5);
return !(distance >= 0x8000'0000ULL && distance <= ~0x8000'0000ULL);
}
inline bool IsWithin2G(const Xbyak::CodeGenerator& code, uintptr_t target) {
return IsWithin2G(reinterpret_cast<uintptr_t>(code.getCurr()), target);
}
template <typename T>
inline void CallFarFunction(Xbyak::CodeGenerator& code, const T f) {
static_assert(std::is_pointer_v<T>, "Argument must be a (function) pointer.");
size_t addr = reinterpret_cast<size_t>(f);
if (IsWithin2G(code, addr)) {
code.call(f);
} else {
// ABI_RETURN is a safe temp register to use before a call
code.mov(ABI_RETURN, addr);
code.call(ABI_RETURN);
}
}
} // namespace Common::X64

46
src/common/x64/xbyak_util.h
View file

@ -1,46 +0,0 @@
// SPDX-FileCopyrightText: 2016 Citra Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#pragma once
#include <type_traits>
#include <xbyak/xbyak.h>
#include "common/x64/xbyak_abi.h"
namespace Common::X64 {
// Constants for use with cmpps/cmpss
enum {
CMP_EQ = 0,
CMP_LT = 1,
CMP_LE = 2,
CMP_UNORD = 3,
CMP_NEQ = 4,
CMP_NLT = 5,
CMP_NLE = 6,
CMP_ORD = 7,
};
constexpr bool IsWithin2G(uintptr_t ref, uintptr_t target) {
const u64 distance = target - (ref + 5);
return !(distance >= 0x8000'0000ULL && distance <= ~0x8000'0000ULL);
}
inline bool IsWithin2G(const Xbyak::CodeGenerator& code, uintptr_t target) {
return IsWithin2G(reinterpret_cast<uintptr_t>(code.getCurr()), target);
}
template <typename T>
inline void CallFarFunction(Xbyak::CodeGenerator& code, const T f) {
static_assert(std::is_pointer_v<T>, "Argument must be a (function) pointer.");
size_t addr = reinterpret_cast<size_t>(f);
if (IsWithin2G(code, addr)) {
code.call(f);
} else {
// ABI_RETURN is a safe temp register to use before a call
code.mov(ABI_RETURN, addr);
code.call(ABI_RETURN);
}
}
} // namespace Common::X64