#include "convolution.h"
#include <mpi.h>
#include <stdio.h>

static float *input, *output, *filter;
static int N, C, H, W;
static int K, R, S;
static int OH, OW;
static int pad;
static int dilation;
static int stride;
static int mpi_rank, mpi_world_size;

// dilation=1,, padding=0, stride=1, node=2
// N, C, K, H, W 2^n, R, S=16

void convolution_opt(){

   for (int n = 0; n < N; ++n) {
      for (int k = 0; k < K; ++k) {
        for (int oh = 0; oh < OH; ++oh) {
          for (int ow = 0; ow < OW; ++ow) {
            float o = 0.f;
            for (int c = 0; c < C; ++c) {
              for (int r = 0; r < R; ++r) {
                for (int s = 0; s < S; ++s) {
                  int h = oh * stride - pad + r * dilation;
                  int w = ow * stride - pad + s * dilation;
                  if (h < 0 || h >= H || w < 0 || w >= W) continue;
                  float i = input[n * C * H * W + c * H * W + h * W + w];
                  float f = filter[k * C * R * S + c * R * S + r * S + s];
                  o += i * f;
                }
              }
            }
            output[n * K * OH * OW + k * OH * OW + oh * OW + ow] = o;
          }
        }
      }
    }

}


void convolution(
    float *_input, float *_output, float *_filter, 
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride) {
  input = _input;
  output = _output;
  filter = _filter;

  int num_threads = 20;

  OH = (H + 2 * pad - dilation * (R - 1) - 1) / stride + 1;
  OW = (W + 2 * pad - dilation * (S - 1) - 1) / stride + 1;
  
  // Node가 1 or 2
  // Case 1 ( Node = 2 )
  // N 기준 분할
  
  

  if (mpi_rank != 0){
	  allc_mat(&input, Ndev, C, H, W);
	  &A, M, K);


  // Master Node
  if(mpi_rank == 0 && mpi_world_size == 2){
	MPI_Isend(&offset, 1, MPI_INT, 1, 0, MPI_COMM_WORLD, &request);
	MPI_Isend(&rows, 1, MPI_INT, 1, 0, mtype, MPI_COMM_WORLD, &request);
	MPI_Isend(&A[offset*K], rows*K, MPI_FLOAT, dest, mtype, MPI_COMM_WORLD, &request);
	MPI_Isend(B, K*N, MPI_FLOAT, dest, mtype, MPI_COMM_WORLD, &request);

	mat_mul_omp();

  }
  
  else if(mpi_rank == 0 && mpi_world_size == 1){

	  // Node 1에서만 계산 (MPI_send 없이)
	  
  }
  else if(mpi_rank == 1 && mpi_world_size == 2){

	// Recv 받아서 Calculate 후 전송

  }


  // Slave Node
  else{
  // 남은 Case 존재 X
  }




}


void convolution_init(
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride) {
  N = _N; C = _C; H = _H; W = _W;
  K = _K; R = _R; S = _S;
  pad = _pad;
  dilation = _dilation;
  stride = _stride;

  MPI_Comm_rank(MPI_COMM_WORLD, &mpi_rank);
  MPI_Comm_size(MPI_COMM_WORLD, &mpi_world_size);

  // Node가 1 or 2
  // Case 1 ( Node = 2 )
  // N 기준 분할
  

  int Ndev = N/mpi_world_size;
  
  // Master Node
  if(mpi_rank == 0 && mpi_world_size == 2){
	   );
  }
  else if(mpi_rank == 0 && mpi_world_size == 1){

	  // Node 1에서만 계산 (MPI_send 없이)
	  
  }
  else if(mpi_rank == 1 && mpi_world_size == 2){

	// Recv 받아서 Calculate 후 전송

  }


  // Slave Node
  else{
  // 남은 Case 존재 X
  }



	  /*
	  for(int i=1; i<mpi_world<size; i++){
		  MPI_Isend(   );
	  }
	  else{
		  MPI_Recv(  );
	  }

	  MPI_Bcast(  );

	  mat_mul_omp();


	  // Gather
	  if (mpi_rank == 0){
		  MPI_Recv( );
	  }
	  else{
		  MPI_Isend( );
	  }
  }*/



}

void convolution_final(
    int _N, int _C, int _H, int _W,
    int _K, int _R, int _S,
    int _pad, int _dilation, int _stride) {
}