chundoong-lab-ta/SamsungDS22/submissions/HW2/junjip.lee/mat_mul.cpp

99 lines
2.1 KiB
C++

#include "mat_mul.h"
//#include <algorithm>
#include <cstdlib>
#include <cstdio>
#include <pthread.h>
#ifndef max
#define max(a,b) (((a) > (b)) ? (a) : (b))
#endif
#ifndef min
#define min(a,b) (((a) < (b)) ? (a) : (b))
#endif
static float *A, *B, *C;
static int M, N, K;
static int num_threads;
struct threadinfo
{
int NumofThread;
int idx;
};
int portion;
static void* mat_mul_thread(void *data) {
// TODO: parallelize & optimize matrix multiplication
threadinfo *thInfo=(threadinfo*)data;
//printf("%s %d %d %d created \n",__func__,__LINE__,thInfo->NumofThread,thInfo->idx);
int totalthread=thInfo->NumofThread;
int idx=thInfo->idx;
int pid = thInfo->idx;
int slice = M / num_threads;
int start = pid * slice;
int end = pid == num_threads - 1 ? M : (pid + 1) * slice;
float Aik;
int bs = 32;
for (int kk = 0; kk < K; kk += bs) {
// for (int jj = 0; jj < N; jj += bs) {
for (int i = start; i < end; ++i) {
for (int k = kk; k < min(kk + bs, K); ++k) {//blocking
// for(int k=0; k<K;++k){
Aik = A[i * K + k];
// for (int j = jj; j < min(jj + bs, N); ++j) {//blocking
for (int j = 0; j < N; ++j){
C[i * N + j] += Aik * 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;
// TODO: create '_num_threads' pthreads
num_threads = _num_threads;
threadinfo *thInfo[81];//thread structure;
for(int i=0;i<num_threads;i++)
{
thInfo[i]= (threadinfo *)malloc(sizeof(threadinfo));
thInfo[i]->NumofThread = num_threads;
}
pthread_t thread[81];
if(M % num_threads==0)
{
portion=M/num_threads;
}
else{
portion=M/num_threads+1;
}
for(int i=0;i<num_threads;i++)
{
thInfo[i]->idx =i;
pthread_create(&thread[i], NULL, mat_mul_thread, (void *)thInfo[i]);
}
for(int i=0;i<num_threads;i++)
pthread_join(thread[i], NULL);
for(int i=0;i<num_threads;i++)
free(thInfo[i]);
}