Autodetect number of threads (remove setting). Fix some bugs.

Author: hrydgard

Common/CPUDetect.h

@@ -39,7 +39,10 @@ struct CPUInfo {
 	bool Mode64bit;
 
 	bool HTT;
+
+	// Number of real CPU cores.
 	int num_cores;
+	// Number of logical CPUs per core.
 	int logical_cpu_count;
 
 	bool bAtom;

Common/Thread/Channel.h

@@ -2,11 +2,18 @@
 
 #include <mutex>
 #include <condition_variable>
+#include <cassert>
+
+// Named Channel.h because I originally intended to support a multi item channel as
+// well as a simple blocking mailbox. Let's see if we get there.
 
 // Single item mailbox.
 template<class T>
 struct Mailbox {
 	Mailbox() : refcount_(1) {}
+	~Mailbox() {
+		assert(refcount_ == 0);
+	}
 
 	std::mutex mutex_;
 	std::condition_variable condvar_;

Common/Thread/ParallelLoop.cpp

@@ -1,6 +1,7 @@
 #include <cstring>
 
-#include "ParallelLoop.h"
+#include "Common/Thread/ParallelLoop.h"
+#include "Common/CPUDetect.h"
 
 class LoopRangeTask : public Task {
 public:
@@ -24,7 +25,6 @@ WaitableCounter *ParallelRangeLoopWaitable(ThreadManager *threadMan, const std::
 		minSize = 1;
 	}
 
-	// TODO: Optimize using minSize.
 	int numTasks = threadMan->GetNumLooperThreads();
 
 	int range = upper - lower;
@@ -37,7 +37,7 @@ WaitableCounter *ParallelRangeLoopWaitable(ThreadManager *threadMan, const std::
 		// Just assign one task per thread, as many as we have.
 		WaitableCounter *counter = new WaitableCounter(range);
 		for (int i = 0; i < range; i++) {
-			threadMan->EnqueueTaskOnThread(i, new LoopRangeTask(counter, loop, i, i + 1));
+			threadMan->EnqueueTaskOnThread(i, new LoopRangeTask(counter, loop, i, i + 1), TaskType::CPU_COMPUTE);
 		}
 		return counter;
 	} else {
@@ -50,14 +50,21 @@ WaitableCounter *ParallelRangeLoopWaitable(ThreadManager *threadMan, const std::
 			int start = lastEnd;
 			d += dx;
 			int end = i == numTasks - 1 ? range : (int)d;
-			threadMan->EnqueueTaskOnThread(i, new LoopRangeTask(counter, loop, start, end));
+			threadMan->EnqueueTaskOnThread(i, new LoopRangeTask(counter, loop, start, end), TaskType::CPU_COMPUTE);
 			lastEnd = end;
 		}
 		return counter;
 	}
 }
 
 void ParallelRangeLoop(ThreadManager *threadMan, const std::function<void(int, int)> &loop, int lower, int upper, int minSize) {
+	if (cpu_info.num_cores == 1) {
+		// "Optimization" for single-core devices.
+		// No point in adding threading overhead, let's just do it inline.
+		loop(lower, upper);
+		return;
+	}
+
 	if (minSize == -1) {
 		minSize = 4;
 	}

Common/Thread/Promise.h

@@ -10,13 +10,11 @@ class PromiseTask : public Task {
 public:
 	PromiseTask() {}
 	~PromiseTask() {
-		tx_->Release();
 	}
 
 	void Run() override {
 		T *value = fun_();
 		tx_->Send(value);
-		tx_->Release();
 	}
 
 	std::function<T *()> fun_;
@@ -30,18 +28,17 @@ class PromiseTask : public Task {
 template<class T>
 class Promise {
 public:
-	static Promise<T> *Spawn(ThreadManager *threadman, std::function<T *()> fun) {
+	static Promise<T> *Spawn(ThreadManager *threadman, std::function<T *()> fun, TaskType taskType) {
 		// std::pair<Rx<T>, Tx<T>> channel = CreateChannel<T>();
 		Mailbox<T> *mailbox = new Mailbox<T>();
 
 		PromiseTask<T> *task = new PromiseTask<T>();
 		task->fun_ = fun;
 		task->tx_ = mailbox;
-		threadman->EnqueueTask(task);
+		threadman->EnqueueTask(task, taskType);
 
 		Promise<T> *promise = new Promise<T>();
 		promise->rx_ = mailbox;
-		mailbox->AddRef();
 		return promise;
 	}
 

Common/Thread/ThreadManager.cpp

@@ -11,10 +11,24 @@
 #include "Common/Thread/ThreadUtil.h"
 #include "Common/Thread/ThreadManager.h"
 
+// Threads and task scheduling
+//
+// * The threadpool should contain a number of threads that's the the number of cores,
+//   plus a fixed number more for I/O-limited background tasks.
+// * Parallel compute-limited loops should use as many threads as there are cores.
+//   They should always be scheduled to the first N threads.
+// * For some tasks, splitting the input values up linearly between the threads
+//   is not fair. However, we ignore that for now.
+
+const int MAX_CORES_TO_USE = 16;
+const int EXTRA_THREADS = 4;  // For I/O limited tasks
+
 struct GlobalThreadContext {
 	std::mutex mutex; // associated with each respective condition variable
 	std::deque<Task *> queue;
 	std::vector<ThreadContext *> threads_;
+
+	int roundRobin;
 };
 
 struct ThreadContext {
@@ -80,7 +94,11 @@ static void WorkerThreadFunc(GlobalThreadContext *global, ThreadContext *thread)
 	}
 }
 
-void ThreadManager::Init(int numThreads) {
+void ThreadManager::Init(int numRealCores, int numLogicalCores) {
+	numComputeThreads_ = std::min(numRealCores, MAX_CORES_TO_USE);
+	int numThreads = numComputeThreads_ + EXTRA_THREADS;
+	numThreads_ = numThreads;
+
 	for (int i = 0; i < numThreads; i++) {
 		ThreadContext *thread = new ThreadContext();
 		thread->cancelled.store(false);
@@ -91,9 +109,25 @@ void ThreadManager::Init(int numThreads) {
 }
 
 void ThreadManager::EnqueueTask(Task *task, TaskType taskType) {
+	int maxThread;
+	int threadOffset = 0;
+	if (taskType == TaskType::CPU_COMPUTE) {
+		// only the threads reserved for heavy compute.
+		maxThread = numComputeThreads_;
+		threadOffset = 0;
+	} else {
+		// any free thread
+		maxThread = numThreads_;
+		threadOffset = numComputeThreads_;
+	}
+
 	// Find a thread with no outstanding work.
-	for (int i = 0; i < global_->threads_.size(); i++) {
-		ThreadContext *thread = global_->threads_[i];
+	int threadNum = threadOffset;
+	for (int i = 0; i < maxThread; i++, threadNum++) {
+		if (threadNum >= global_->threads_.size()) {
+			threadNum = 0;
+		}
+		ThreadContext *thread = global_->threads_[threadNum];
 		if (thread->queueSize.load() == 0) {
 			std::unique_lock<std::mutex> lock(thread->mutex);
 			thread->private_queue.push_back(task);
@@ -104,11 +138,13 @@ void ThreadManager::EnqueueTask(Task *task, TaskType taskType) {
 		}
 	}
 
-	// Still not scheduled? Put it on the global queue and notify a random thread.
+	// Still not scheduled? Put it on the global queue and notify a thread chosen by round-robin.
+	// Not particularly scientific, but hopefully we should not run into this too much.
 	{
 		std::unique_lock<std::mutex> lock(global_->mutex);
 		global_->queue.push_back(task);
-		global_->threads_[0]->cond.notify_one();
+		global_->threads_[global_->roundRobin % maxThread]->cond.notify_one();
+		global_->roundRobin++;
 	}
 }
 
@@ -123,10 +159,7 @@ void ThreadManager::EnqueueTaskOnThread(int threadNum, Task *task, TaskType task
 }
 
 int ThreadManager::GetNumLooperThreads() const {
-	// If possible, let's use all threads but one for parallel loops.
-	// Not sure what's the best policy here. Maybe we should just have more threads than CPUs.
-	int numLooperThreads = (int)(global_->threads_.size()) - 1;
-	return std::max(numLooperThreads, 1);
+	return numComputeThreads_;
 }
 
 void ThreadManager::TryCancelTask(uint64_t taskID) {

Common/Thread/ThreadManager.h

@@ -41,9 +41,12 @@ class ThreadManager {
 	ThreadManager();
 	~ThreadManager();
 
-	void Init(int numWorkerThreads);
-	void EnqueueTask(Task *task, TaskType taskType = TaskType::CPU_COMPUTE);
-	void EnqueueTaskOnThread(int threadNum, Task *task, TaskType taskType = TaskType::CPU_COMPUTE);
+	// The distinction here is to be able to take hyper-threading into account.
+	// It gets even trickier when you think about mobile chips with BIG/LITTLE, but we'll
+	// just ignore it and let the OS handle it.
+	void Init(int numRealCores, int numLogicalCores);
+	void EnqueueTask(Task *task, TaskType taskType);
+	void EnqueueTaskOnThread(int threadNum, Task *task, TaskType taskType);
 
 	// Currently does nothing. It will always be best-effort - maybe it cancels,
 	// maybe it doesn't. Note that the id is the id() returned by the task. You need to make that
@@ -57,6 +60,9 @@ class ThreadManager {
 private:
 	GlobalThreadContext *global_;
 
+	int numThreads_ = 0;
+	int numComputeThreads_ = 0;
+
 	friend struct ThreadContext;
 };
 

Common/Thread/ThreadUtil.cpp

@@ -1,3 +1,5 @@
+#include "ppsspp_config.h"
+
 #ifdef _WIN32
 #include <windows.h>
 #ifdef __MINGW32__
@@ -16,6 +18,9 @@
 
 #if defined(__ANDROID__) || defined(__APPLE__) || (defined(__GLIBC__) && defined(_GNU_SOURCE))
 #include <pthread.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/syscall.h>
 #endif
 
 #ifdef TLS_SUPPORTED
@@ -41,12 +46,14 @@ static EXCEPTION_DISPOSITION NTAPI ignore_handler(EXCEPTION_RECORD *rec,
 #endif
 
 void SetCurrentThreadName(const char* threadName) {
-#ifdef _WIN32
+#if PPSSPP_PLATFORM(WINDOWS)
 	// Set the debugger-visible threadname through an unholy magic hack
 	static const DWORD MS_VC_EXCEPTION = 0x406D1388;
 #endif
 
-#if defined(_WIN32) && defined(__MINGW32__)
+	// TODO: Use the new function SetThreadDescription available since Windows 10, version 1607.
+
+#if PPSSPP_PLATFORM(WINDOWS) && defined(__MINGW32__)
 	// Thread information for VS compatible debugger. -1 sets current thread.
 	THREADNAME_INFO ti;
 	ti.dwType = 0x1000;
@@ -67,7 +74,7 @@ void SetCurrentThreadName(const char* threadName) {
 
 	// Pop exception handler
 	tib->ExceptionList = tib->ExceptionList->Next;
-#elif defined(_WIN32)
+#elif PPSSPP_PLATFORM(WINDOWS)
 #pragma pack(push,8)
 	struct THREADNAME_INFO {
 		DWORD dwType; // must be 0x1000
@@ -121,3 +128,11 @@ void AssertCurrentThreadName(const char *threadName) {
 	}
 #endif
 }
+
+int GetCurrentThreadIdForDebug() {
+#if PPSSPP_PLATFORM(WINDOWS)
+	return (int)GetCurrentThreadId();
+#else
+	return gettid();
+#endif
+}

Common/Thread/ThreadUtil.h

@@ -6,3 +6,7 @@
 // for AssertCurrentThreadName to work.
 void SetCurrentThreadName(const char *threadName);
 void AssertCurrentThreadName(const char *threadName);
+
+// Just gets a cheap thread identifier so that you can see different threads in debug output,
+// exactly what it is is badly specified and not useful for anything.
+int GetCurrentThreadIdForDebug();

Core/Config.cpp

@@ -423,12 +423,6 @@ std::string CreateRandMAC() {
 	return randStream.str();
 }
 
-static int DefaultNumWorkers() {
-	// Let's cap the global thread pool at 16 threads. Nothing we do really should have much
-	// use for more...
-	return std::min(16, cpu_info.num_cores);
-}
-
 static int DefaultCpuCore() {
 #if PPSSPP_ARCH(ARM) || PPSSPP_ARCH(ARM64) || PPSSPP_ARCH(X86) || PPSSPP_ARCH(AMD64)
 	return (int)CPUCore::JIT;
@@ -473,7 +467,6 @@ static ConfigSetting generalSettings[] = {
 	ConfigSetting("DiscordPresence", &g_Config.bDiscordPresence, true, true, false),  // Or maybe it makes sense to have it per-game? Race conditions abound...
 	ConfigSetting("UISound", &g_Config.bUISound, false, true, false),
 
-	ReportedConfigSetting("NumWorkerThreads", &g_Config.iNumWorkerThreads, &DefaultNumWorkers, true, true),
 	ConfigSetting("AutoLoadSaveState", &g_Config.iAutoLoadSaveState, 0, true, true),
 	ReportedConfigSetting("EnableCheats", &g_Config.bEnableCheats, false, true, true),
 	ConfigSetting("CwCheatRefreshRate", &g_Config.iCwCheatRefreshRate, 77, true, true),

Core/Config.h

@@ -72,7 +72,6 @@ struct Config {
 	bool bBrowse; // when opening the emulator, immediately show a file browser
 
 	// General
-	int iNumWorkerThreads;
 	bool bScreenshotsAsPNG;
 	bool bUseFFV1;
 	bool bDumpFrames;