void alloc_tex()
{
int i, ow = tex_w, oh = tex_h;
for (tex_w = 1; tex_w < width; tex_w <<= 1);
for (tex_h = 1; tex_h < height; tex_h <<= 1);
if (tex_h != oh || tex_w != ow) {
tex = realloc(tex, tex_h * tex_w * 3 + tex_h * sizeof(rgb_t*));
free_shared(SHARED_tex);
SHARED_tex = malloc_shared(MIN(tex_h * tex_w * 3, BLOCK_SIZE), tex + tex_h);
}
for (tex[0] = (rgb_t *)(tex + tex_h), i = 1; i < tex_h; i++)
tex[i] = tex[i - 1] + tex_w;
}
int main() {
int num_threads = 4;
int global[num_threads];
int compare[num_threads];
int increment[num_threads];
pthread_t pool[num_threads];
struct args arg_array[num_threads];
int i, j, max = 10;
//printf("Incrementing by %d for %d iterations:\n", increment, max);
int *SHARED_MEM_global = malloc_shared(num_threads*sizeof(int), &global, LEGUP_RAM_LOCATION_ONCHIP);
for (j = 0; j<num_threads;j++) {
global[j] = 5;
compare[j] = 5;
increment[j] = 2*j;
}
for (i = 0; i<max; i++) {
// pre-accelerator copy
memcpy_to_shared(SHARED_MEM_global, &global, num_threads*sizeof(int));
// accelerator call
for (j = 0; j < num_threads; j++) {
struct args tmp = {&SHARED_MEM_global[j], increment[j]};
memcpy(arg_array+j, &tmp, sizeof(tmp));
pthread_create(pool+j, NULL, ByRef_thread, (void *)(arg_array + j));
}
for (j = 0; j < num_threads; j++) {
pthread_join(pool[j], NULL);
}
//ByRef(SHARED_MEM_global, increment);
// post-accelerator copy
memcpy_from_shared(&global, SHARED_MEM_global, num_threads*sizeof(int));
for (j = 0; j < num_threads; j++) {
printf("value from thread %d is %d\n", j, global[j]);
}
}
free_shared(SHARED_MEM_global);
return 0;
}
