#include <libkern/OSAtomic.h>

#include "context.h"
#include "origin_cs.h"

static void start_worker(Worker* worker);

union Data* createCPUWorker(struct Context* context, int id, Queue* queue) {
    struct Worker* worker = ALLOC(context, Worker);
    struct CPUWorker* cpuWorker = ALLOC(context, 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 (union Data*)(worker);
}

static void start_worker(Worker* worker) {
    CPUWorker* cpuWorker = (CPUWorker*)worker->worker;
    cpuWorker->context = NEW(struct Context);
    initContext(cpuWorker->context);
    pthread_cond_wait(&cpuWorker->cond, &cpuWorker->mutex);
    goto meta(cpuWorker->context, C_taskReceiveWorker);
}

__code taskReceiveWorker(struct Context* context, Worker* worker) {
    Queue* queue = worker->tasks;
    queue->next = C_getTask1;
    goto meta(context, queue->take);
}

__code taskReceiveWorker_stub(struct Context* context) {
    goto taskReceiveWorker(context, Gearef(context, Worker));
}

__code getTask1(struct Context* context, Worker* worker, struct Context* task) {
    if (!task)
        return; // end thread
    task->worker = worker;
    goto meta(task, task->next);
}

__code getTask1_stub(struct Context* context) {
    Worker* worker = Gearef(context,Worker);
    Context* task = &worker->tasks->data->context; 
    goto getTask1(context,worker,task);
}

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

__code shutdownWorker(struct Context* context, CPUWorker* worker) {
}

__code shutdownWorker_stub(struct Context* context) {
    CPUWorker* worker = (CPUWorker *)GearImpl(context, Worker, worker);
    goto shutdownWorker(context,worker);
}