#include "mat_mul.h"

#include <cstdlib>
#include <cstdio>
#include <pthread.h>

static float *A, *B, *C;
static int M, N, K;
static int num_threads;

int min(int a,int b)
{
	return (a<b)?a:b;
}
static void* mat_mul_thread(void *data) {
  // TODO: parallelize & optimize matrix multiplication
  int index = * (int *) data;

  int offset = M / num_threads;
  int start = index * offset;
  int end = (index == num_threads - 1) ? M : (index + 1) * offset;
  int kk, i, k, j;

  float A_calc;
  int bs = 32;

  for (kk = 0; kk < K; kk += bs) {
  	for (i = start; i < end; ++i) {
        	for (k = kk; k < min(kk + bs, K); k++) {
          		A_calc = A[i * K + k];
			if(!(N%8)){
          			for (j = 0; j < N; j=j+4) {
	    				C[i * N + j] += A_calc * B[k * N + j];
            				C[i * N + j+1] += A_calc * B[k * N + j+1];
            				C[i * N + j+2] += A_calc * B[k * N + j+2];
            				C[i * N + j+3] += A_calc * B[k * N + j+3];
            
           				//C[i * N + j+4] += A_calc * B[k * N + j+4];
           	 			//C[i * N + j+5] += A_calc * B[k * N + j+5];
            				//C[i * N + j+6] += A_calc * B[k * N + j+6];
            				//C[i * N + j+7] += A_calc * B[k * N + j+7];
				}
			}
			else{
				for(j=0;j<N;j++){
					C[i * N + j] += A_calc * B[k * N + j];
				}
			}
		}
	}
  }
  return NULL;
}

void mat_mul(float *_A, float *_B, float *_C, int _M, int _N, int _K, int _num_threads) {
  A = _A, B = _B, C = _C;
  M = _M, N = _N, K = _K;
  num_threads = _num_threads;

  // TODO: create '_num_threads' pthreads
  pthread_t thread[num_threads];
  for(int i=0;i<num_threads;i++){
	int* index = (int*)malloc(sizeof(int));
      	*index=i;
  	pthread_create(&thread[i], NULL, mat_mul_thread, index);
  }
  for(int i=0;i<num_threads;i++)
	  pthread_join(thread[i], NULL);
}