[core/core_timing] better MWAITX and WAITPKG delays (#3984)

This implements MWAITX and WAITPKG extensions (umonitor, mwait) for CPUs that support them.

Reduces wait times and bypasses the timing stuff from the OS that is slow (windows notably). generally it should answer within 0.2 to 0.5 microsecs (since most requests wait for that long).

Also does a general rework of static ctors and stuff

Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3984
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: crueter <crueter@eden-emu.dev>
This commit is contained in:
lizzie
2026-05-30 21:59:10 +02:00
committed by crueter
parent ff7bbaea7d
commit 7c32cf03a1
19 changed files with 477 additions and 533 deletions
+10 -28
View File
@@ -8,15 +8,13 @@
#include <mutex>
#include <string>
#include <tuple>
#include "common/cpu_features.h"
#include "common/cpu_features.h"
#ifdef _WIN32
#include "common/windows/timer_resolution.h"
#endif
#ifdef ARCHITECTURE_x86_64
#include "common/x64/cpu_wait.h"
#endif
#include "common/settings.h"
#include "core/core_timing.h"
#include "core/hardware_properties.h"
@@ -47,8 +45,7 @@ struct CoreTiming::Event {
}
};
CoreTiming::CoreTiming() : clock{Common::CreateOptimalClock()} {}
CoreTiming::CoreTiming() = default;
CoreTiming::~CoreTiming() {
Reset();
}
@@ -64,31 +61,16 @@ void CoreTiming::Initialize(std::function<void()>&& on_thread_init_) {
Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
on_thread_init();
has_started = true;
// base frequency in MHz: 1ns (10^-9) = 1GHz (10^9)
while (!stop_token.stop_requested()) {
while (!paused && !stop_token.stop_requested()) {
paused_set = false;
if (auto const next_time = Advance(); next_time) {
// There are more events left in the queue, wait until the next event.
auto wait_time = *next_time - GetGlobalTimeNs().count();
auto const wait_time = *next_time - GetGlobalTimeNs().count();
if (wait_time > 0) {
#ifdef _WIN32
while (!paused && !event.IsSet() && wait_time > 0) {
wait_time = *next_time - GetGlobalTimeNs().count();
if (wait_time >= timer_resolution_ns) {
Common::Windows::SleepForOneTick();
} else {
#ifdef ARCHITECTURE_x86_64
Common::X64::MicroSleep();
#else
std::this_thread::yield();
#endif
}
}
if (event.IsSet())
event.Reset();
#else
event.WaitFor(std::chrono::nanoseconds(wait_time));
#endif
}
} else {
// Queue is empty, wait until another event is scheduled and signals us to
@@ -226,7 +208,7 @@ void CoreTiming::ResetTicks() {
}
u64 CoreTiming::GetClockTicks() const {
u64 fres = is_multicore ? clock.GetCNTPCT() : Common::WallClock::CPUTickToCNTPCT(cpu_ticks);
u64 fres = is_multicore ? Common::g_wall_clock.GetCNTPCT() : Common::WallClock::CPUTickToCNTPCT(cpu_ticks);
if (auto const overclock = Settings::values.fast_cpu_time.GetValue(); overclock != Settings::CpuClock::Off) {
fres = u64(f64(fres) * (1.7 + 0.3 * u32(overclock)));
}
@@ -240,7 +222,7 @@ u64 CoreTiming::GetClockTicks() const {
u64 CoreTiming::GetGPUTicks() const {
return is_multicore
? clock.GetGPUTick()
? Common::g_wall_clock.GetGPUTick()
: Common::WallClock::CPUTickToGPUTick(cpu_ticks);
}
@@ -317,14 +299,14 @@ void CoreTiming::Reset() {
/// @brief Returns current time in nanoseconds.
std::chrono::nanoseconds CoreTiming::GetGlobalTimeNs() const noexcept {
return is_multicore
? clock.GetTimeNS()
? Common::g_wall_clock.GetTimeNS()
: std::chrono::nanoseconds{Common::WallClock::CPUTickToNS(cpu_ticks)};
}
/// @brief Returns current time in microseconds.
std::chrono::microseconds CoreTiming::GetGlobalTimeUs() const noexcept {
return is_multicore
? clock.GetTimeUS()
? Common::g_wall_clock.GetTimeUS()
: std::chrono::microseconds{Common::WallClock::CPUTickToUS(cpu_ticks)};
}