[chore] fix build errors

Signed-off-by: Caio Oliveira <caiooliveirafarias0@gmail.com>

src/common/thread.cpp

@@ -5,9 +5,11 @@
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <string>
+#include <thread>
 
 #include "common/error.h"
 #include "common/logging/log.h"
+#include "common/assert.h"
 #include "common/thread.h"
 #ifdef __APPLE__
 #include <mach/mach.h>
@@ -18,6 +20,8 @@
 #include "common/string_util.h"
 #else
 #if defined(__Bitrig__) || defined(__DragonFly__) || defined(__FreeBSD__) || defined(__OpenBSD__)
+#include <sys/cpuset.h>
+#include <sys/_cpuset.h>
 #include <pthread_np.h>
 #endif
 #include <pthread.h>
@@ -28,7 +32,7 @@
 #endif
 
 #ifdef __FreeBSD__
-#define cpu_set_t cpuset_t
+#   define cpu_set_t cpuset_t
 #endif
 
 namespace Common {
@@ -77,22 +81,14 @@ void SetCurrentThreadPriority(ThreadPriority new_priority) {
 #endif
 }
 
+void SetCurrentThreadName(const char* name) {
 #ifdef _MSC_VER
-
-// Sets the debugger-visible name of the current thread.
-void SetCurrentThreadName(const char* name) {
-    static auto pf = (decltype(&SetThreadDescription))(void*)GetProcAddress(GetModuleHandle(TEXT("KernelBase.dll")), "SetThreadDescription");
-    if (pf)
+    // Sets the debugger-visible name of the current thread.
+    if (auto pf = (decltype(&SetThreadDescription))(void*)GetProcAddress(GetModuleHandle(TEXT("KernelBase.dll")), "SetThreadDescription"); pf)
         pf(GetCurrentThread(), UTF8ToUTF16W(name).data()); // Windows 10+
-}
-
-#else // !MSVC_VER, so must be POSIX threads
-
-// MinGW with the POSIX threading model does not support pthread_setname_np
-void SetCurrentThreadName(const char* name) {
-    // See for reference
-    // https://gitlab.freedesktop.org/mesa/mesa/-/blame/main/src/util/u_thread.c?ref_type=heads#L75
-#ifdef __APPLE__
+    else
+        ; // No-op
+#elif  defined(__APPLE__)
     pthread_setname_np(name);
 #elif defined(__HAIKU__)
     rename_thread(find_thread(NULL), name);
@@ -112,13 +108,33 @@ void SetCurrentThreadName(const char* name) {
         pthread_setname_np(pthread_self(), buf);
     }
 #elif defined(_WIN32)
-    // mingw stub
+    // MinGW with the POSIX threading model does not support pthread_setname_np
+    // See for reference
+    // https://gitlab.freedesktop.org/mesa/mesa/-/blame/main/src/util/u_thread.c?ref_type=heads#L75
     (void)name;
 #else
     pthread_setname_np(pthread_self(), name);
 #endif
 }
 
+void PinCurrentThreadToPerformanceCore(size_t core_id) {
+    ASSERT(core_id < 4);
+    // If we set a flag for a CPU that doesn't exist, the thread may not be allowed to
+    // run in ANY processor!
+    auto const total_cores = std::thread::hardware_concurrency();
+    if (core_id < total_cores) {
+#if defined(__linux__) || defined(__FreeBSD__)
+        cpu_set_t set;
+        CPU_ZERO(&set);
+        CPU_SET(core_id, &set);
+        pthread_setaffinity_np(pthread_self(), sizeof(set), &set);
+#elif defined(_WIN32)
+        DWORD set = 1UL << core_id;
+        SetThreadAffinityMask(GetCurrentThread(), set);
+#else
+        // No pin functionality implemented
 #endif
+    }
+}
 
 } // namespace Common

src/common/thread.h

@@ -106,7 +106,7 @@ enum class ThreadPriority : u32 {
 };
 
 void SetCurrentThreadPriority(ThreadPriority new_priority);
-
 void SetCurrentThreadName(const char* name);
+void PinCurrentThreadToPerformanceCore(size_t core_id);
 
 } // namespace Common

src/core/cpu_manager.cpp

@@ -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 +186,10 @@ 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);
+    Common::PinCurrentThreadToPerformanceCore(core);
     auto& data = core_data[core];
     data.host_context = Common::Fiber::ThreadToFiber();
 

src/dynarmic/src/dynarmic/backend/x64/block_of_code.cpp

@@ -194,8 +194,6 @@ HostFeature GetHostFeatures() {
         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::tBMI2)) {
         // BMI2 instructions such as pdep and pext have been very slow up until Zen 3.

src/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.cpp.inc

@@ -420,11 +420,10 @@ void AxxEmitX64::EmitExclusiveWriteMemoryInline(AxxEmitContext& ctx, IR::Inst* i
     const Xbyak::Reg64 vaddr = ctx.reg_alloc.UseGpr(code, args[1]);
     const Xbyak::Reg32 status = ctx.reg_alloc.ScratchGpr(code).cvt32();
     const Xbyak::Reg64 tmp = ctx.reg_alloc.ScratchGpr(code);
-    const Xbyak::Reg64 tmp2 = ctx.reg_alloc.ScratchGpr(code);
 
     const auto wrapped_fn = exclusive_write_fallbacks[std::make_tuple(ordered, bitsize, vaddr.getIdx(), value.getIdx())];
 
-    EmitExclusiveLock(code, conf, tmp, tmp2.cvt32());
+    EmitExclusiveLock(code, conf, tmp, eax);
 
     SharedLabel end = GenSharedLabel();
 

src/dynarmic/src/dynarmic/backend/x64/emit_x64_memory.h

@@ -346,7 +346,7 @@ void EmitExclusiveLock(BlockOfCode& code, const UserConfig& conf, Xbyak::Reg64 p
     }
 
     code.mov(pointer, std::bit_cast<u64>(GetExclusiveMonitorLockPointer(conf.global_monitor)));
-    EmitSpinLockLock(code, pointer, tmp, code.HasHostFeature(HostFeature::WAITPKG));
+    EmitSpinLockLock(code, pointer, tmp);
 }
 
 template<typename UserConfig>

src/dynarmic/src/dynarmic/backend/x64/host_feature.h

@@ -35,10 +35,9 @@ enum class HostFeature : u64 {
     BMI2 = 1ULL << 19,
     LZCNT = 1ULL << 20,
     GFNI = 1ULL << 21,
-    WAITPKG = 1ULL << 22,
 
     // Zen-based BMI2
-    FastBMI2 = 1ULL << 23,
+    FastBMI2 = 1ULL << 22,
 
     // Orthographic AVX512 features on 128 and 256 vectors
     AVX512_Ortho = AVX512F | AVX512VL,

src/dynarmic/src/dynarmic/common/spin_lock_x64.cpp

@@ -22,46 +22,17 @@ static const auto default_cg_mode = nullptr; //Allow RWE
 
 namespace Dynarmic {
 
-void EmitSpinLockLock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp, bool waitpkg) {
-    // TODO: this is because we lack regalloc - so better to be safe :(
-    if (waitpkg) {
-        code.push(Xbyak::util::eax);
-        code.push(Xbyak::util::ebx);
-        code.push(Xbyak::util::edx);
-    }
+void EmitSpinLockLock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp) {
     Xbyak::Label start, loop;
+
     code.jmp(start, code.T_NEAR);
     code.L(loop);
-    if (waitpkg) {
-        // TODO: This clobbers EAX and EDX did we tell the regalloc?
-        // ARM ptr for address-monitoring
-        code.umonitor(ptr);
-        // tmp.bit[0] = 0: C0.1 | Slow Wakup | Better Savings
-        // tmp.bit[0] = 1: C0.2 | Fast Wakup | Lesser Savings
-        // edx:eax is implicitly used as a 64-bit deadline timestamp
-        // Use the maximum so that we use the operating system's maximum
-        // allowed wait time within the IA32_UMWAIT_CONTROL register
-        // Enter power state designated by tmp and wait for a write to lock_ptr
-        code.mov(Xbyak::util::eax, 0xFFFFFFFF);
-        code.mov(Xbyak::util::edx, Xbyak::util::eax);
-        // TODO: We can only be here because tmp is 1 already - however we repeatedly overwrite it...
-        code.mov(Xbyak::util::ebx, 1);
-        code.umwait(Xbyak::util::ebx);
-        // CF == 1 if we hit the OS-timeout in IA32_UMWAIT_CONTROL without a write
-        // CF == 0 if we exited the wait for any other reason
-    } else {
-        code.pause();
-    }
+    code.pause();
     code.L(start);
     code.mov(tmp, 1);
     /*code.lock();*/ code.xchg(code.dword[ptr], tmp);
     code.test(tmp, tmp);
     code.jnz(loop, code.T_NEAR);
-    if (waitpkg) {
-        code.pop(Xbyak::util::edx);
-        code.pop(Xbyak::util::ebx);
-        code.pop(Xbyak::util::eax);
-    }
 }
 
 void EmitSpinLockUnlock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp) {
@@ -89,7 +60,7 @@ void SpinLockImpl::Initialize() noexcept {
     Xbyak::Reg64 const ABI_PARAM1 = Backend::X64::HostLocToReg64(Backend::X64::ABI_PARAM1);
     code.align();
     lock = code.getCurr<void (*)(volatile int*)>();
-    EmitSpinLockLock(code, ABI_PARAM1, code.eax, false);
+    EmitSpinLockLock(code, ABI_PARAM1, code.eax);
     code.ret();
     code.align();
     unlock = code.getCurr<void (*)(volatile int*)>();

src/dynarmic/src/dynarmic/common/spin_lock_x64.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 /* This file is part of the dynarmic project.
  * Copyright (c) 2022 MerryMage
  * SPDX-License-Identifier: 0BSD
@@ -12,7 +9,7 @@
 
 namespace Dynarmic {
 
-void EmitSpinLockLock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp, bool waitpkg);
+void EmitSpinLockLock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp);
 void EmitSpinLockUnlock(Xbyak::CodeGenerator& code, Xbyak::Reg64 ptr, Xbyak::Reg32 tmp);
 
 }  // namespace Dynarmic