Plexi09/eden-miror
1
0
Fork 0
mirror of https://git.eden-emu.dev/eden-emu/eden synced 2026-04-10 20:38:57 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

[android, core] pin core threads to logical CPUs 0-3 (#3121)

Browse source 
this basically allows the threads to exist in these logical CPUs, undisturbed, and without trashing each other's cache
this could improve performance, very tricky thing to pull off correctly, but again, this is mostly an experiment
will mainly benefit: Linux, Android, FreeBSD, Windows (not ARM)
Additionally, this means no context trashing :)

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

Co-authored-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3121
Reviewed-by: DraVee <dravee@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
This commit is contained in:
lizzie 2026-02-01 01:54:59 +01:00 • committed by crueter
parent acf7deea95
commit 1925726b96
No known key found for this signature in database
GPG key ID: 425ACD2D4830EBC6
3 changed files with 51 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

23
src/core/cpu_manager.cpp
View file

@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
@ -7,6 +7,7 @@
#include "common/fiber.h"
#include "common/scope_exit.h"
#include "common/thread.h"
#include "common/settings.h"
#include "core/core.h"
#include "core/core_timing.h"
#include "core/cpu_manager.h"
@ -25,11 +26,8 @@ CpuManager::~CpuManager() = default;
void CpuManager::Initialize() {
num_cores = is_multicore ? Core::Hardware::NUM_CPU_CORES : 1;
gpu_barrier = std::make_unique<Common::Barrier>(num_cores + 1);
for (std::size_t core = 0; core < num_cores; core++) {
core_data[core].host_thread =
std::jthread([this, core](std::stop_token token) { RunThread(token, core); });
}
for (std::size_t core = 0; core < num_cores; core++)
core_data[core].host_thread = std::jthread([this, core](std::stop_token token) { RunThread(token, core); });
}
void CpuManager::Shutdown() {
@ -188,14 +186,15 @@ void CpuManager::ShutdownThread() {
void CpuManager::RunThread(std::stop_token token, std::size_t core) {
/// Initialization
system.RegisterCoreThread(core);
std::string name;
if (is_multicore) {
name = "CPUCore_" + std::to_string(core);
} else {
name = "CPUThread";
}
std::string name = is_multicore ? ("CPUCore_" + std::to_string(core)) : std::string{"CPUThread"};
Common::SetCurrentThreadName(name.c_str());
Common::SetCurrentThreadPriority(Common::ThreadPriority::Critical);
#ifdef __ANDROID__
// Aimed specifically for Snapdragon 8 Elite devices
// This kills performance on desktop, but boosts perf for UMA devices
// like the S8E. Mediatek and Mali likely won't suffer.
Common::PinCurrentThreadToPerformanceCore(core);
#endif
auto& data = core_data[core];
data.host_context = Common::Fiber::ThreadToFiber();