95 lines
2.5 KiB
C++
95 lines
2.5 KiB
C++
|
#include "mat_mul.h"
|
||
|
|
#include "util.h"
|
||
|
|
#include <cstdio>
|
||
|
|
#include <cstdlib>
|
||
|
|
#include <mpi.h>
|
||
|
|
#include "omp.h"
|
||
|
|
|
||
|
|
static float *A, *B, *C;
|
||
|
|
static int M, N, K;
|
||
|
|
static int num_threads;
|
||
|
|
static int mpi_rank, mpi_world_size;
|
||
|
|
|
||
|
|
static int min(int x, int y) {
|
||
|
|
return x < y ? x : y;
|
||
|
|
}
|
||
|
|
|
||
|
|
static void mat_mul_omp() {
|
||
|
|
#define ITILESIZE (32)
|
||
|
|
#define JTILESIZE (1024)
|
||
|
|
#define KTILESIZE (1024)
|
||
|
|
|
||
|
|
omp_set_num_threads(num_threads);
|
||
|
|
|
||
|
|
#pragma omp parallel for
|
||
|
|
for (int ii = 0; ii < M; ii += ITILESIZE) {
|
||
|
|
for (int jj = 0; jj < N; jj += JTILESIZE) {
|
||
|
|
for (int kk = 0; kk < K; kk += KTILESIZE) {
|
||
|
|
|
||
|
|
for (int k = kk; k < min(K, kk + KTILESIZE); k++) {
|
||
|
|
for (int i = ii; i < min(M, ii + ITILESIZE); i++) {
|
||
|
|
float ar = A[i * K + k];
|
||
|
|
for (int j = jj; j < min(N, jj + JTILESIZE); j+=1) {
|
||
|
|
C[i * N + j] += ar * B[k * N + j];
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
void mat_mul(float *_A, float *_B, float *_C, int _M, int _N, int _K,
|
||
|
|
int _num_threads, int _mpi_rank, int _mpi_world_size) {
|
||
|
|
A = _A, B = _B, C = _C;
|
||
|
|
M = _M, N = _N, K = _K;
|
||
|
|
num_threads = _num_threads, mpi_rank = _mpi_rank,
|
||
|
|
mpi_world_size = _mpi_world_size;
|
||
|
|
|
||
|
|
MPI_Status status;
|
||
|
|
MPI_Request req;
|
||
|
|
int averow = M / mpi_world_size;
|
||
|
|
int extra = M % mpi_world_size;
|
||
|
|
|
||
|
|
if (mpi_rank == 0) {
|
||
|
|
//timer_start(0);
|
||
|
|
|
||
|
|
for (int node = 1, offset = averow + extra; node < mpi_world_size; node++) {
|
||
|
|
//printf("Master sending partial matrix A to node(%d)\n", mpi_rank);
|
||
|
|
MPI_Isend(B, K * N, MPI_FLOAT, node, node, MPI_COMM_WORLD, &req);
|
||
|
|
MPI_Isend(&A[offset * K], averow * K, MPI_FLOAT, node, node, MPI_COMM_WORLD, &req);
|
||
|
|
offset += averow;
|
||
|
|
}
|
||
|
|
|
||
|
|
//printf("Sending time %f\n", timer_stop(0));
|
||
|
|
M = averow + extra;
|
||
|
|
mat_mul_omp();
|
||
|
|
}
|
||
|
|
else {
|
||
|
|
alloc_mat(&B, K, N);
|
||
|
|
MPI_Recv(B, K * N, MPI_FLOAT, 0, mpi_rank, MPI_COMM_WORLD, &status);
|
||
|
|
alloc_mat(&A, averow, K);
|
||
|
|
alloc_mat(&C, averow, N);
|
||
|
|
zero_mat(C, averow, N);
|
||
|
|
|
||
|
|
MPI_Recv(A, averow * K, MPI_FLOAT, 0, mpi_rank, MPI_COMM_WORLD, &status);
|
||
|
|
//printf("Worker node(%d) received partial matrix A\n", mpi_rank);
|
||
|
|
|
||
|
|
M = averow;
|
||
|
|
mat_mul_omp();
|
||
|
|
|
||
|
|
MPI_Isend(C, averow * N, MPI_FLOAT, 0, mpi_rank, MPI_COMM_WORLD, &req);
|
||
|
|
//printf("Worker node(%d) sent calculation result to Master\n", mpi_rank);
|
||
|
|
}
|
||
|
|
|
||
|
|
if (mpi_rank == 0) {
|
||
|
|
//timer_start(1);
|
||
|
|
for (int node = 1, offset = averow + extra; node < mpi_world_size; node++) {
|
||
|
|
MPI_Recv(&C[offset * N], averow * N, MPI_FLOAT, node, node, MPI_COMM_WORLD, &status);
|
||
|
|
offset += averow;
|
||
|
|
}
|
||
|
|
//printf("Receiving time %f\n", timer_stop(1));
|
||
|
|
}
|
||
|
|
}
|