#include <starpu.h>
#include <stdlib.h>
#include <stdio.h>
#include <fpga.h>
#include <starpu_scheduler.h>
#include "../helper.h"

#include "StreamFMA.max"
#include "MaxSLiCInterface.h"
#include "StreamFMA.h"

#define SIZE 128

static max_engine_t *engine ;
static max_actions_t*act;
static max_file_t *maxfile;

void cpu_func(void *buffers[], void *cl_arg)
{
    int *a = (int*) STARPU_VECTOR_GET_PTR(buffers[0]);
    int *b = (int*) STARPU_VECTOR_GET_PTR(buffers[1]);
    int *c = (int*) STARPU_VECTOR_GET_PTR(buffers[2]);

    int size = STARPU_VECTOR_GET_NX(buffers[0]);

    (void)buffers;
    (void)cl_arg;

    int i;
    for (i = 0; i < size; i++)
	c[i] = a[i] + b[i];
}

void fpga_mult(void *buffers[], void *cl_arg)
{   
    (void)cl_arg;
    
    int *a = (int*) STARPU_VECTOR_GET_PTR(buffers[0]);
    int *b = (int*) STARPU_VECTOR_GET_PTR(buffers[1]);
    int *c = (int*) STARPU_VECTOR_GET_PTR(buffers[2]);

    int size = STARPU_VECTOR_GET_NX(buffers[0]);

    //Actions to run on an engine
     act = max_actions_init(maxfile, NULL);

    //set the number of ticks for a kernel
    max_set_ticks  (act, "StreamFMAKernel", size);

#if 1
// sera remplacé par le transfert fait par le copy_fpga_to_ram/ram_to_fpga
    //add data to an input stream
    //max_set_param_uint64t(act, "address", 0);
    //max_set_param_uint64t(act, "nbytes", size *sizeof(a[0]));
    max_queue_input(act, "a", a, size *sizeof(a[0]));
    
    max_queue_input(act, "b", b, size*sizeof(b[0]));
    max_queue_output(act,"output", c, size*sizeof(c[0]));

// et à la place, ici on récupère le pointeur dans la mémoire FPGA avec STARPU_VECTOR_GET_PTR(descr[0]), et c'est ça qu'on donne à l'implémentation fpga
#endif

    //run actions on the engine
    printf("Running on DFE using dynamic interface ...\n");

    printf("**** Run actions in non blocking mode **** \n");

    //run actions in non_blocking mode
    max_run_t *run0= max_run_nonblock(engine, act);

    printf("*** wait for the actions on DFE to complete *** \n");
    //wait for the actions to complete
    max_wait(run0);
    
  }

static struct starpu_codelet cl =
{
    .cpu_funcs = {cpu_func},
    .cpu_funcs_name = {"cpu_func"},
//#ifdef STARPU_USE_FPGA
    .fpga_funcs = {fpga_mult},
    .fpga_funcs_name={"fpga_mult"},
//#endif
    .nbuffers = 3,
    .modes = {STARPU_R, STARPU_R, STARPU_W}
};


int main(int argc, char **argv)
{

    /* Enable profiling */
    starpu_profiling_status_set(1);

    struct starpu_conf conf;
    starpu_data_handle_t handle_a, handle_b, handle_c;
    int ret;
    int size=1234;

    starpu_conf_init(&conf);

    conf.sched_policy_name = "eager";
    conf.calibrate = 0;

    ret = starpu_initialize(&conf, &argc, &argv);
    if (ret == -ENODEV) return STARPU_TEST_SKIPPED;
    STARPU_CHECK_RETURN_VALUE(ret, "starpu_init");

    //Implementation of a maxfile
     maxfile = StreamFMA_init();

    //Implementation of an engine
    engine = max_load(maxfile, "*");

    int a[SIZE];
    int b[SIZE];
    int c[SIZE];

    int i;
    for(i = 0; i < SIZE; ++i)
    {
        a[i] = random() % 100;
        b[i] = random() % 100;
    }

    starpu_vector_data_register(&handle_a, STARPU_MAIN_RAM, (uintptr_t) &a, SIZE, sizeof(int));
    starpu_vector_data_register(&handle_b, STARPU_MAIN_RAM, (uintptr_t) &b, SIZE, sizeof(int));
    starpu_vector_data_register(&handle_c, STARPU_MAIN_RAM, (uintptr_t) &c, SIZE, sizeof(int));

    struct starpu_task *task = starpu_task_create();
    task->cl = &cl;
    task->handles[0] = handle_a;
    task->handles[1] = handle_b;
    task->handles[2] = handle_c;
    
    task->synchronous = 1;
    task->destroy = 0;
    /* submit the task to StarPU */

    //starpu_task_destroy(task);
    ret = starpu_task_submit(task);
    
    fprintf(stderr,"task submitted %d\n", ret);

    starpu_data_unregister(handle_a);
    starpu_data_unregister(handle_b);
    starpu_data_unregister(handle_c);
  
	int mysize = SIZE;
	if (mysize > 10)
		mysize = 10;
	for (i = 0; i < mysize; ++i) 
	{
		printf("%d == %d\n", c[i], a[i] + b[i]);
	}

#if 1
// -> main
    //deallocate the set of actions
    max_actions_free(act);

    //unload and deallocate an engine obtained by way of max_load
    max_unload(engine);
#endif

    starpu_shutdown();

    return EXIT_SUCCESS;

}