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
89
90
91
92
93
94
95
96
97
|
#include "ParDiagonalMatrix.h"
using MPI_Wrappers::allgatherv;
namespace ATC_matrix {
// template<>
// void ParDiagonalMatrix<double>::MultAB(const Matrix<double> &B, DenseMatrix<double> &C) const
// {
// //SparseMatrix<T>::compress(*this);
// GCK(*this, B, this->nCols()!=B.nRows(), "ParDiagonalMatrix * Matrix");
// const INDEX nRows = this->nRows();
// const INDEX nCols = this->nCols();
// // Determine which rows will be handled on this processor
// int nProcs = MPI_Wrappers::size(_comm);
// int myRank = MPI_Wrappers::rank(_comm);
// INDEX startIndex = (myRank * nRows) / nProcs;
// INDEX endIndex = ((myRank + 1) * nRows) / nProcs;
// // Calculate the scaled rows associated with this processor
// for (INDEX i = startIndex; i < endIndex; i++) {
// double value = (*this)[i];
// for (INDEX j = 0; j < nCols; j++)
// C(i, j) = value * B(i, j);
// }
// // Collect results on all processors
// // consider sending only owned rows from each processor
// allsum(_comm, MPI_IN_PLACE, C.ptr(), C.size());
// }
template<>
void ParDiagonalMatrix<double>::MultAB(const Matrix<double> &B, DenseMatrix<double> &C) const
{
//SparseMatrix<T>::compress(*this);
GCK(*this, B, this->nCols()!=B.nRows(), "ParDiagonalMatrix * Matrix");
const INDEX nRows = this->nRows();
const INDEX nCols = this->nCols();
int nProcs = MPI_Wrappers::size(_comm);
int myRank = MPI_Wrappers::rank(_comm);
#ifdef COL_STORAGE // Column-major storage
int nMajor = nCols;
int nMinor = nRows;
#else // Row-major storage
int nMajor = nRows;
int nMinor = nCols;
#endif
int *majorCounts = new int[nProcs];
int *majorOffsets = new int[nProcs];
// Determine which rows/columns will be handled on this processor
for (int i = 0; i < nProcs; i++) {
majorOffsets[i] = (i * nMajor) / nProcs;
majorCounts[i] = (((i + 1) * nMajor) / nProcs) - majorOffsets[i];
}
INDEX myNMajor = majorCounts[myRank];
INDEX myMajorOffset = majorOffsets[myRank];
// Calculate the scaled values associated with this processor, in row chunks
#ifdef COL_STORAGE // Column-major storage
for (INDEX i = 0; i < nRows; i++) {
double value = (*this)[i];
for (INDEX j = myMajorOffset; j < myMajorOffset + myNMajor; j++)
C(i, j) = value * B(i, j);
}
#else // Row-major storage
for (INDEX i = myMajorOffset; i < myMajorOffset + myNMajor; i++) {
double value = (*this)[i];
for (INDEX j = 0; j < nCols; j++)
C(i, j) = value * B(i, j);
}
#endif
for (int i = 0; i < nProcs; i++) {
majorCounts[i] *= nMinor;
majorOffsets[i] *= nMinor;
}
// Collect results on all processors
allgatherv(_comm, C.ptr() + myMajorOffset * nMinor, myNMajor * nMinor,
C.ptr(), majorCounts, majorOffsets);
delete[] majorCounts;
delete[] majorOffsets;
}
} // end namespace
|
