Cannon矩阵乘法的MPI实现和性能分析.doc

#include stdio.h #include mpi.h #include math.h #include stdlib.h #include time.h #include memory.h MPI_Status status; double **A, **B, **C; //C=A*B double *a,*b,*c; //各个进程的缓冲区 int n; //矩阵的行列数 int np; //每个进程控制的小矩阵的行列数 int p,rank; //进程个个数、当前进程的编号,笛卡尔进程编号 double *tempa, *tempb; void ProduceABC(); //在根处理器中生成矩阵AB初始化矩阵C void PrintABC();//输出结果 void ScatterAB();// 分发矩阵AB中的元素到各个进程中 void MainProcess(); //cannon算法的主过程 void collectC(); //收集结果矩阵C void Mutiply(); //矩阵相乘 void Printab(); void Printc(); int main(int argc, char *argv[]) { int i; double starttime,endtime; MPI_Init(argc, argv); MPI_Comm_size(MPI_COMM_WORLD, p); MPI_Comm_rank(MPI_COMM_WORLD, rank); if(rank == 0) { printf(请输入矩阵的行列数n= ); fflush(stdout); scanf_s(%d, n); printf(\n); } MPI_Bcast(n, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD); // n = atoi(argv[1]); np = n/(int)sqrt(p); a = (double*)malloc(np*np*sizeof(double)); b = (double*)malloc(np*np*sizeof(double)); c = (double*)malloc(np*np*sizeof(double)); memset(c, 0, np*np*sizeof(double)); tempa = (double*)malloc(np*np*sizeof(double)); tempb = (double*)malloc(np*np*sizeof(double)); if(rank == 0) { //在根处理器中为矩阵ABC分配空间 A = (double**)malloc(n*sizeof(double*)); B = (double**)malloc(n*sizeof(double*)); C = (double**)malloc(n*sizeof(double*)); for(i = 0; i n; i++) { A[i] = (double*)malloc(n*sizeof(double)); B[i] = (double*)malloc(n*sizeof(double)); C[i] = (double*)malloc(n*sizeof(double)); } ProduceABC(); //在根处理器中随机生成矩阵AB，初始化矩阵C ScatterAB();// 分发矩阵AB中的元素到各个进程中 } else { MPI_Recv(a, np*np, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD, status); MPI_Recv(b, np*np, MPI_DOUBLE, 0, 2, MPI_COMM_WORLD, status); } starttime=MPI_Wtime(); //开始时间 MainProcess(); //cannon算法的主过程 if(rank == 0) { collectC(); //收集结果矩阵C PrintABC(); //输出结果 endtime=MPI_Wtime(); printf(time used: %lf\n,endtime - starttime); for(i = 0; i n; i++) { free(A[i]); free(B[i]); free(C[i]); } free(A); free(B); free(C); } else {