Plexi09/eden-miror
1
0
Fork 0
mirror of https://git.eden-emu.dev/eden-emu/eden synced 2026-06-03 20:27:08 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

[core/core_timing] better MWAITX and WAITPKG delays

Browse source 
Signed-off-by: lizzie <lizzie@eden-emu.dev>
This commit is contained in:
lizzie 2026-05-19 07:46:59 +00:00
parent 9d55875377
commit 394578d038
3 changed files with 32 additions and 55 deletions
Download patch file Download diff file Expand all files Collapse all files

81
src/common/x64/cpu_wait.cpp
View file

@ -5,6 +5,8 @@
#ifdef _MSC_VER #ifdef _MSC_VER
#include <intrin.h> #include <intrin.h>
#else
#include <x86intrin.h>
#endif #endif
#include "common/x64/cpu_detect.h" #include "common/x64/cpu_detect.h"
@ -13,60 +15,33 @@
namespace Common::X64 { namespace Common::X64 {
namespace { #ifdef __clang__
__attribute__((target("waitpkg")))
// 100,000 cycles is a reasonable amount of time to wait to save on CPU resources. __attribute__((target("mwaitx")))
// For reference: #elif defined(__GNUC__)
// At 1 GHz, 100K cycles is 100us #pragma GCC target("waitpkg")
// At 2 GHz, 100K cycles is 50us #pragma GCC target("mwaitx")
// At 4 GHz, 100K cycles is 25us
constexpr auto PauseCycles = 100'000U;
} // Anonymous namespace
#if defined(_MSC_VER) && !defined(__clang__)
__forceinline static void TPAUSE() {
static constexpr auto RequestC02State = 0U;
_tpause(RequestC02State, FencedRDTSC() + PauseCycles);
}
__forceinline static void MWAITX() {
static constexpr auto EnableWaitTimeFlag = 1U << 1;
static constexpr auto RequestC1State = 0U;
// monitor_var should be aligned to a cache line.
alignas(64) u64 monitor_var{};
_mm_monitorx(&monitor_var, 0, 0);
_mm_mwaitx(EnableWaitTimeFlag, RequestC1State, PauseCycles);
}
#else
static void TPAUSE() {
static constexpr auto RequestC02State = 0U;
const auto tsc = FencedRDTSC() + PauseCycles;
const auto eax = static_cast<u32>(tsc & 0xFFFFFFFF);
const auto edx = static_cast<u32>(tsc >> 32);
asm volatile("tpause %0" : : "r"(RequestC02State), "d"(edx), "a"(eax));
}
static void MWAITX() {
static constexpr auto EnableWaitTimeFlag = 1U << 1;
static constexpr auto RequestC1State = 0U;
// monitor_var should be aligned to a cache line.
alignas(64) u64 monitor_var{};
asm volatile("monitorx" : : "a"(&monitor_var), "c"(0), "d"(0));
asm volatile("mwaitx" : : "a"(RequestC1State), "b"(PauseCycles), "c"(EnableWaitTimeFlag));
}
#endif #endif
void MicroSleep(u64 rem) {
void MicroSleep() { // 100,000 cycles is a reasonable amount of time to wait to save on CPU resources.
static const bool has_waitpkg = GetCPUCaps().waitpkg; // For reference:
static const bool has_monitorx = GetCPUCaps().monitorx; // At 1 GHz, 100K cycles is 100us
// At 2 GHz, 100K cycles is 50us
if (has_waitpkg) { // At 4 GHz, 100K cycles is 25us
TPAUSE(); auto& caps = GetCPUCaps();
} else if (has_monitorx) { u32 cycles = caps.invariant_tsc
MWAITX(); ? 1'000'000U
: rem * (caps.tsc_frequency / 1000000ULL);
if (caps.waitpkg) {
constexpr auto RequestC02State = 0U;
_tpause(RequestC02State, FencedRDTSC() + cycles);
} else if (caps.monitorx) {
constexpr auto EnableWaitTimeFlag = 1U << 1;
constexpr auto RequestC1State = 0U;
// monitor_var should be aligned to a cache line.
alignas(64) u64 monitor_var{};
_mm_monitorx(&monitor_var, 0, 0);
_mm_mwaitx(EnableWaitTimeFlag, RequestC1State, cycles);
} else { } else {
std::this_thread::yield(); std::this_thread::yield();
} }

4
src/common/x64/cpu_wait.h
View file

@ -3,8 +3,10 @@
#pragma once #pragma once
#include "common/common_types.h"
namespace Common::X64 { namespace Common::X64 {
void MicroSleep(); void MicroSleep(u64 rem);
} // namespace Common::X64 } // namespace Common::X64

2
src/core/core_timing.cpp
View file

@ -78,7 +78,7 @@ void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) {
Common::Windows::SleepForOneTick(); Common::Windows::SleepForOneTick();
} else { } else {
#ifdef ARCHITECTURE_x86_64 #ifdef ARCHITECTURE_x86_64
Common::X64::MicroSleep(); Common::X64::MicroSleep(wait_time);
#else #else
std::this_thread::yield(); std::this_thread::yield();
#endif #endif