1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
| void InitProcess (double* &A,double* &B,double* &C ,int &Size) {
MPI_Comm_size(MPI_COMM_WORLD, &ProcNum);
MPI_Comm_rank(MPI_COMM_WORLD, &ProcRank);
if (ProcRank == 0) {
do {printf("n–matrix multiplication–");
Size = 4;
if (Size< ProcNum) printf("Matrix size is less than the number of processes! n");
if (Size%ProcNum!=0) printf("Matrix size should be dividable by the number of processes! n");
}
while ((Size< ProcNum)||(Size%ProcNum!=0));
}
if (Size<3) flag=1;
MPI_Bcast(&Size, 1, MPI_INT, 0, MPI_COMM_WORLD);
if (ProcRank == 0) {
A = new double [Size*Size];
B = new double [Size*Size];
C = new double [Size*Size];
RandInit (A, Size); RandInit (B, Size);//Инициализация матриц
}
}
//————————————————-
void Flip (double *&B, int dim) {
double temp=0.0;
for (int i=0;i<Size;i++){
for (int j=i+1;j<Size;j++)
temp=B[i*dim+j],
B[i*dim+j]=B[j*dim+i],
B[j*dim+i]=temp;
}
}
//————————————————-
void MatrixMultiplicationMPI(double *&A, double *&B, double *&C, int &Size) {
int dim =Size;
int i, j, k, p, ind;
double temp;
MPI_Status Status;
int ProcPartSize=dim/ProcNum;//Количество подзадач
int ProcPartElem=ProcPartSize*dim;//Количество элементов в подзадаче
double* bufA=new double[dim*ProcPartSize];
double* bufB=new double[dim*ProcPartSize];
double* bufC=new double[dim*ProcPartSize];
int ProcPart = dim/ProcNum, part = ProcPart*dim;
if (ProcRank == 0) {
Flip(B,Size);//Транспонирование матрицы В
}
MPI_Scatter(A, part, MPI_DOUBLE, bufA, part, MPI_DOUBLE, 0, MPI_COMM_WORLD);
MPI_Scatter(B, part, MPI_DOUBLE, bufB, part, MPI_DOUBLE, 0, MPI_COMM_WORLD);
temp=0.0;
for(i=0;i<ProcPartSize;i++){
for(j=0;j<ProcPartSize;j++){
for(k=0;k<dim;k++)
temp += bufA[i*dim+k]*bufB[j*dim+k],
bufC[i*dim+j+ProcPartSize*ProcRank]=temp,
temp=0.0;
}
}
int NextProc; int PrevProc;
for(p=1;p<ProcNum;p++){
NextProc = ProcRank+1;
if ( ProcRank ==ProcNum-1 )
NextProc = 0;
PrevProc = ProcRank-1;
if ( ProcRank == 0 )
PrevProc = ProcNum-1;
MPI_Sendrecv_replace(bufB,part, MPI_DOUBLE, NextProc, 0, PrevProc, 0, MPI_COMM_WORLD, &Status);//Сдвиг вдоль процессов
temp=0.0;
for (i=0; i<ProcPartSize;i++){
for(j=0;j<ProcPartSize; j++){
for(k=0;k<dim;k++){
temp +=bufA[i*dim+k]*bufB[j*dim+k];}
if (ProcRank-p >=0 )
ind=ProcRank-p;
else ind=(ProcNum-p+ProcRank);
bufC[i*dim+j+ind*ProcPartSize]=temp;
temp=0.0;
}
}
}
MPI_Gather(bufC, ProcPartElem, MPI_DOUBLE, C, ProcPartElem, MPI_DOUBLE, 0, MPI_COMM_WORLD);
delete []bufA;
delete []bufB;
delete []bufC;
} |
(
