#include <libkern/OSAtomic.h>

#include "../context.h"

static void start_worker(Worker* worker);

Worker* createCPUWorker(struct Context* context, int id, Queue* queue, enum Code next) {
    struct Worker* worker = new Worker();
    struct CPUWorker* cpuWorker = new CPUWorker();
    worker->worker = (union Data*)cpuWorker;
    worker->tasks = queue;
    cpuWorker->id = id;
    worker->taskReceive = C_taskReceiveWorker;
    worker->shutdown = C_shutdownWorker;
    pthread_create(&worker->worker->CPUWorker.thread, NULL, (void*)&start_worker, worker);
    return worker;
}

static void start_worker(Worker* worker) {
    CPUWorker* cpuWorker = (CPUWorker*)worker->worker;
    cpuWorker->context = NEW(struct Context);
    initContext(cpuWorker->context);
    Gearef(cpuWorker->context, Worker)->worker = (union Data*)worker;
    goto meta(cpuWorker->context, C_taskReceiveWorker);
}

__code taskReceiveWorker(struct Worker* worker,struct Queue* queue) {
    queue->queue = (union Data*)worker->tasks;
    queue->next = C_getTask;
    goto meta(context, worker->tasks->take);
}

__code taskReceiveWorker_stub(struct Context* context) {
    CPUWorker* cpuWorker = (CPUWorker *)GearImpl(context, Worker, worker);
    pthread_cond_wait(&cpuWorker->cond, &cpuWorker->mutex);
    goto taskReceiveWorker(context, &Gearef(context, Worker)->worker->Worker, Gearef(context, Queue));
}

__code getTask(struct Worker* worker, struct Context* task) {
    if (!task)
        return; // end thread
    task->worker = worker;
    context->next = C_taskReceiveWorker; // set CG after task exec
    goto meta(task, task->next);
}

__code getTask_stub(struct Context* context) {
    Worker* worker = &Gearef(context,Worker)->worker->Worker;
    struct Context* task = &Gearef(context, Queue)->data->Context;
    goto getTask(context, worker, task);
}

#ifdef USE_CUDA
__code twiceGpu() {
    cuMemcpyHtoDAsync(context,context,context,context->stream);
    cuLaunchkanel();
    cuMemcpyDtoHAsync();
}
#endif

__code shutdownWorker(struct CPUWorker* worker) {
}