我有这个矩阵添加m程序。当我使用一个小的矩阵文件时，它工作正常，但当它与大矩阵文件（5300 * 4200）一起使用时，它会在处理器高于2时崩溃。错误是“内存不足”。
#include<mpi.h>
#include<stdio.h>
#include<stdlib.h>
#include<ctype.h>
#include<time.h>
int main(int argc, char** argv)
{
char buffer[256];
FILE* pIn = NULL;
FILE* pIn1 = NULL;
FILE* pOut = NULL;
char inputF[128];
char inputF1[128];
char outF[128];
int rowA = 0;
int columnA = 0;
int rowB = 0;
int columnB = 0;
int averow = 0;
int rows = 0;
int extra = 0,offset = 0;
double **mat;
double **mat2;
double **sum;
double **sum2;
double StartTime = 0;
double EndTime = 0;
int id,p;
int i = 0,j = 0,source =0,dest =0;
MPI_Init(&argc,&argv);
MPI_Comm_rank(MPI_COMM_WORLD,&id);
MPI_Comm_size(MPI_COMM_WORLD,&p);
if(id == 0 || p == 1)
{
/////////OPEN FILE AND READ ROWS AND COLUMNS////////////////////
fgets(buffer,sizeof(buffer),stdin);
sscanf(buffer,"%s",inputF);
fgets(buffer,sizeof(buffer),stdin);
sscanf(buffer,"%s",inputF1);
//printf("File A: %s File B: %s\n",inputF,inputF1);
pIn = fopen_s(inputF,"r");
if(pIn == NULL)
{
printf("Error opening File");
exit(0);
}
//printf("File 1 reading Success!\n");
fgets(buffer,200,pIn);
sscanf(buffer,"%d %d",&rowA,&columnA);
pIn1 = fopen(inputF1,"r");
if(pIn1 == NULL)
{
printf("Error opening File");
exit(0);
}
//printf("File 2 reading Success!\n");
fgets(buffer,200,pIn1);
sscanf(buffer,"%d %d",&rowB,&columnB);
//printf("Dimensions of two matrices A: %d %d B: %d %d \n",rowA,columnA,rowB,columnB);
/////////OPEN FILE AND READ ROWS AND COLUMNS////////////////////
///////////ALLOCATING MEMORY FOR 3 MATRICES/////////////////////
mat = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
mat[i] = (double*)calloc(columnA,sizeof(double));
if(mat == NULL)
{
printf("Error allocating memory for 1st matrix.");
exit(1);
}
mat2 = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
mat2[i] = (double*)calloc(columnA,sizeof(double));
if(mat2 == NULL)
{
printf("Error allocating memory for 2nd matrix.");
exit(1);
}
sum2 = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
sum2[i] = (double*)calloc(columnA,sizeof(double));
if(sum2 == NULL)
{
printf("Error allocating memory for ANS matrix.");
exit(1);
}
///////////ALLOCATING MEMORY FOR 3 MATRICES/////////////////////
//////READ MATRIX DATA///////////////////////////
for (i = 0; i < rowA; i++)
{
for (j = 0; j < columnA; j++)
{
fscanf(pIn,"%lf",&mat[i][j]);
fscanf(pIn1,"%lf",&mat2[i][j]);
}
}
//////READ MATRIX DATA///////////////////////////
////////DIVIDE WORK AMONG SLAVES///////////////////
if(p > 1)
{
StartTime = MPI_Wtime();
averow = rowA/(p-1);
extra = rowA%(p-1);
offset = 0;
for (dest=1; dest<=(p-1); dest++)
{
rows = (dest <= extra) ? averow+1 : averow;
//printf("Process %d:Sending %d rows to process %d offset=%d\n",id,rows,dest,offset);
MPI_Send(&columnA, 1, MPI_INT, dest, 1, MPI_COMM_WORLD);
MPI_Send(&rowA, 1, MPI_INT, dest, 1, MPI_COMM_WORLD);
MPI_Send(&offset, 1, MPI_INT, dest, 1, MPI_COMM_WORLD);
MPI_Send(&rows, 1, MPI_INT, dest, 1, MPI_COMM_WORLD);
for (i = offset; i < rows+offset; i++)
for (j = 0; j < columnA; j++)
{
MPI_Send(&mat[i][j],5, MPI_DOUBLE, dest, 1,MPI_COMM_WORLD);
MPI_Send(&mat2[i][j],5, MPI_DOUBLE, dest, 1, MPI_COMM_WORLD);
}
offset = offset + rows;
}
////////DIVIDE WORK AMONG SLAVES///////////////////
////////RECIEVE BACK THE ADDED WORK///////////////////
for (source = 1;source <=(p-1);source++)
{
MPI_Recv(&rows, 1, MPI_INT, source, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&offset, 1, MPI_INT, source, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&columnA, 1, MPI_INT, source, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
for (i = offset; i < rows+offset; i++)
for (j = 0; j < columnA; j++)
MPI_Recv(&sum2[i][j], 5, MPI_DOUBLE, source, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
printf("Process %d: Received results from process %d\n",id,source);
}
EndTime = MPI_Wtime();
printf("Timings : %f Sec", EndTime - StartTime);
////////RECIEVE BACK THE ADDED WORK///////////////////
}
////////ADD FOR 1 PROCESSOR///////////////////
if(p == 1)
{
StartTime = MPI_Wtime();
for (i = 0; i < rowA; i++)
for (j = 0; j < columnA; j++)
sum2[i][j] = mat[i][j]+mat2[i][j];
EndTime = MPI_Wtime();
printf("Timings : %f Sec", EndTime - StartTime);
}
////////ADD FOR 1 PROCESSOR///////////////////
/////////////////PRINT RESULTS////////////////////
//for (int i = 0; i < rowA; i++)
//for (int j = 0; j < columnA; j++)
//printf(" %.1lf",sum[i][j]);
/////////////////PRINT RESULTS////////////////////
//////////////WRITE RESULTS TO FILE//////////////
pOut = fopen("out.txt","w");
if(pOut == NULL)
{
printf("Error opening File");
exit(0);
}
fprintf(pOut,"%d %d\n",rowA,columnA);
for (i = 0; i < rowA; i++)
for (j = 0; j < columnA; j++)
fprintf(pOut," %.1lf ",sum2[i][j]);
fclose(pOut);
fclose(pIn);
fclose(pIn1);
//////////////WRITE RESULTS TO FILE//////////////
/////////////FREE MEMORY ALLOCATED///////////////
/////////////FREE MEMORY ALLOCATED///////////////
}
if(id > 0)
{
MPI_Recv(&columnA, 1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&rowA, 1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
///////////ALLOCATING MEMORY FOR 3 MATRICES/////////////////////
mat = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
mat[i] = (double*)calloc(columnA,sizeof(double));
if(mat == NULL)
{
printf("Process %d:Error allocating memory for 1st matrix.",id);
exit(1);
}
mat2 = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
mat2[i] = (double*)calloc(columnA,sizeof(double));
if(mat2 == NULL)
{
printf("Process %d: Error allocating memory for 2nd matrix.",id);
exit(1);
}
sum = (double **)calloc(rowA,sizeof(double*));
for(i = 0; i < rowA; i++)
sum[i] = (double*)calloc(columnA,sizeof(double));
if(sum == NULL)
{
printf("Process %d: Error allocating memory for ANS matrix.",id);
exit(1);
}
///////////ALLOCATING MEMORY FOR 3 MATRICES/////////////////////
//////////RECIEVE AND ADD EACH PART AND SEND BACK FOR EACH SLAVES/////////////
MPI_Recv(&offset, 1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&rows, 1, MPI_INT, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
//printf("Process %d: Recieved this number of Rows: %d\n",id,rows);
for (i = offset; i < rows+offset; i++)
for (j = 0; j < columnA; j++)
{
MPI_Recv(&mat[i][j], 5, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&mat2[i][j],5, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
sum[i][j] = mat[i][j] + mat2[i][j];
//printf(" %d:%.1lf",id,sum[i][j]);
}
//printf("send matrix %.1lf and %.1lf from Process\n",mat[0][0],mat[2][2],id );
MPI_Send(&rows, 1, MPI_INT, 0, 1, MPI_COMM_WORLD);
MPI_Send(&offset, 1, MPI_INT, 0, 1, MPI_COMM_WORLD);
MPI_Send(&columnA, 1, MPI_INT, 0, 1, MPI_COMM_WORLD);
for (i = offset; i < rows+offset; i++)
for (j = 0; j < columnA; j++)
MPI_Send(&sum[i][j],5, MPI_DOUBLE, 0, 1, MPI_COMM_WORLD);
//////////RECIEVE AND ADD AND SEND BACK TO MASTER EACH PART FOR EACH SLAVES/////////////
}
MPI_Finalize();
}