// worker_queue.cpp #define WORKER_TASK_PROC(name) isize name(void *data) typedef WORKER_TASK_PROC(WorkerTaskProc); struct WorkerTask { WorkerTaskProc *do_work; void *data; isize result; }; struct ThreadPool { BlockingMutex mutex; Semaphore sem_available; std::atomic processing_work_count; std::atomic is_running; gbAllocator allocator; MPMCQueue tasks; Thread *threads; isize thread_count; char worker_prefix[10]; i32 worker_prefix_len; }; void thread_pool_init(ThreadPool *pool, gbAllocator const &a, isize thread_count, char const *worker_prefix = nullptr); void thread_pool_destroy(ThreadPool *pool); void thread_pool_start(ThreadPool *pool); void thread_pool_join(ThreadPool *pool); void thread_pool_add_task(ThreadPool *pool, WorkerTaskProc *proc, void *data); THREAD_PROC(worker_thread_internal); void thread_pool_init(ThreadPool *pool, gbAllocator const &a, isize thread_count, char const *worker_prefix) { pool->allocator = a; mpmc_init(&pool->tasks, a, 1024); pool->thread_count = gb_max(thread_count, 0); pool->threads = gb_alloc_array(a, Thread, pool->thread_count); mutex_init(&pool->mutex); semaphore_init(&pool->sem_available); pool->is_running = true; pool->worker_prefix_len = 0; if (worker_prefix) { i32 worker_prefix_len = cast(i32)gb_strlen(worker_prefix); worker_prefix_len = gb_min(worker_prefix_len, 10); gb_memmove(pool->worker_prefix, worker_prefix, worker_prefix_len); pool->worker_prefix_len = worker_prefix_len; } for (isize i = 0; i < pool->thread_count; i++) { Thread *t = &pool->threads[i]; thread_init(t); t->user_index = i; #if 0 // TODO(bill): Fix this on Linux as it causes a seg-fault if (pool->worker_prefix_len > 0) { char worker_name[16] = {}; gb_snprintf(worker_name, gb_size_of(worker_name), "%.*s%u", pool->worker_prefix_len, pool->worker_prefix, cast(u16)i); thread_set_name(t, worker_name); } #endif } } void thread_pool_start(ThreadPool *pool) { for (isize i = 0; i < pool->thread_count; i++) { Thread *t = &pool->threads[i]; thread_start(t, worker_thread_internal, pool); } } void thread_pool_join(ThreadPool *pool) { pool->is_running.store(false); semaphore_post(&pool->sem_available, cast(i32)pool->thread_count); yield(); for (isize i = 0; i < pool->thread_count; i++) { Thread *t = &pool->threads[i]; thread_join(t); } } void thread_pool_destroy(ThreadPool *pool) { thread_pool_join(pool); semaphore_destroy(&pool->sem_available); mutex_destroy(&pool->mutex); gb_free(pool->allocator, pool->threads); pool->thread_count = 0; mpmc_destroy(&pool->tasks); } void thread_pool_add_task(ThreadPool *pool, WorkerTaskProc *proc, void *data) { mutex_lock(&pool->mutex); WorkerTask task = {}; task.do_work = proc; task.data = data; mpmc_enqueue(&pool->tasks, task); semaphore_post(&pool->sem_available, 1); mutex_unlock(&pool->mutex); } bool thread_pool_try_and_pop_task(ThreadPool *pool, WorkerTask *task) { bool got_task = false; if (mpmc_dequeue(&pool->tasks, task)) { pool->processing_work_count.fetch_add(1); got_task = true; } return got_task; } void thread_pool_do_work(ThreadPool *pool, WorkerTask *task) { task->result = task->do_work(task->data); pool->processing_work_count.fetch_sub(1); } void thread_pool_wait_to_process(ThreadPool *pool) { if (pool->thread_count == 0) { WorkerTask task = {}; while (thread_pool_try_and_pop_task(pool, &task)) { thread_pool_do_work(pool, &task); } return; } while (pool->tasks.count.load(std::memory_order_relaxed) > 0 || pool->processing_work_count.load() != 0) { WorkerTask task = {}; if (thread_pool_try_and_pop_task(pool, &task)) { thread_pool_do_work(pool, &task); } // Safety-kick while (pool->tasks.count.load(std::memory_order_relaxed) > 0 && pool->processing_work_count.load() == 0) { mutex_lock(&pool->mutex); semaphore_post(&pool->sem_available, cast(i32)pool->tasks.count.load(std::memory_order_relaxed)); mutex_unlock(&pool->mutex); } yield(); } thread_pool_join(pool); } THREAD_PROC(worker_thread_internal) { ThreadPool *pool = cast(ThreadPool *)thread->user_data; while (pool->is_running.load()) { semaphore_wait(&pool->sem_available); WorkerTask task = {}; if (thread_pool_try_and_pop_task(pool, &task)) { thread_pool_do_work(pool, &task); } } // Cascade semaphore_release(&pool->sem_available); return 0; }