mirror of
https://github.com/OPM/opm-simulators.git
synced 2025-02-25 18:55:30 -06:00
85 lines
3.0 KiB
C++
85 lines
3.0 KiB
C++
|
/*
|
||
|
|
Copyright 2020 Equinor ASA
|
||
|
|
|
||
|
|
This file is part of the Open Porous Media project (OPM).
|
||
|
|
|
||
|
|
OPM is free software: you can redistribute it and/or modify
|
||
|
|
it under the terms of the GNU General Public License as published by
|
||
|
|
the Free Software Foundation, either version 3 of the License, or
|
||
|
|
(at your option) any later version.
|
||
|
|
|
||
|
|
OPM is distributed in the hope that it will be useful,
|
||
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||
|
|
GNU General Public License for more details.
|
||
|
|
|
||
|
|
You should have received a copy of the GNU General Public License
|
||
|
|
along with OPM. If not, see <http://www.gnu.org/licenses/>.
|
||
|
|
*/
|
||
|
|
|
||
|
|
#ifndef FPGA_MATRIX_HEADER_INCLUDED
|
||
|
|
#define FPGA_MATRIX_HEADER_INCLUDED
|
||
|
|
|
||
|
|
#include <vector>
|
||
|
|
|
||
|
|
namespace bda
|
||
|
|
{
|
||
|
|
|
||
|
|
/// This struct resembles a csr matrix, only doubles are supported
|
||
|
|
/// The data is stored in contiguous memory, such that they can be copied to a device in one transfer.
|
||
|
|
class Matrix {
|
||
|
|
|
||
|
|
public:
|
||
|
|
|
||
|
|
/// Allocate Matrix and data arrays with given sizes
|
||
|
|
/// \param[in] N number of rows
|
||
|
|
/// \param[in] nnzs number of nonzeros
|
||
|
|
Matrix(int N_, int nnzs_)
|
||
|
|
: nnzValues(new double[nnzs_]),
|
||
|
|
colIndices(new int[nnzs_]),
|
||
|
|
rowPointers(new int[N_+1]),
|
||
|
|
N(N_),
|
||
|
|
nnzs(nnzs_)
|
||
|
|
{}
|
||
|
|
|
||
|
|
/// All constructors allocate new memory, so always delete here
|
||
|
|
~Matrix(){
|
||
|
|
delete[] nnzValues;
|
||
|
|
delete[] colIndices;
|
||
|
|
delete[] rowPointers;
|
||
|
|
}
|
||
|
|
|
||
|
|
|
||
|
|
/// Converts this matrix to the dataformat used by the FPGA.
|
||
|
|
/// The FPGA uses a new data format called CSRO (Compressed Sparse Row Offset).
|
||
|
|
/// The purpose of this format is to allow the data to be streamable.
|
||
|
|
/// The rowPointers array has an unpredictable reading pattern/timing,
|
||
|
|
/// it also needs a extra work if a row is shorter than a cacheline.
|
||
|
|
/// The array of N+1 rowPointers is replaced by an array of nnz rowOffsets.
|
||
|
|
/// The value of this offset is 0, unless the corresponding nnz is the first of a row,
|
||
|
|
/// in that case it is 'the number of empty rows preceeding it + 1'.
|
||
|
|
/// The FPGA can simply add the rowOffset to the current rowIdx to get the new rowIdx.
|
||
|
|
/// Example:
|
||
|
|
/// [1 0 0 3 0] nnzValues [1 3 2 2 1 4 3 4 1]
|
||
|
|
/// [0 2 2 0 1] colIndices [0 3 1 2 4 0 1 2 4]
|
||
|
|
/// [4 0 0 0 0] -> rowPointers [0 2 5 6 6 9]
|
||
|
|
/// [0 0 0 0 0] rowOffsets [1 0 1 0 0 1 2 0 0]
|
||
|
|
/// [0 3 4 0 1]
|
||
|
|
/// The rowOffset is stored in 1 byte, meaning the maximum value is 255.
|
||
|
|
int toRDF(int numColors, std::vector<int>& nodesPerColor,
|
||
|
|
std::vector<std::vector<int> >& colIndicesInColor, int nnzsPerRowLimit,
|
||
|
|
std::vector<std::vector<double> >& ubNnzValues, short int *ubColIndices, int *nnzValsSizes, unsigned char *NROffsets, int *colorSizes);
|
||
|
|
|
||
|
|
double *nnzValues;
|
||
|
|
int *colIndices;
|
||
|
|
int *rowPointers;
|
||
|
|
int N;
|
||
|
|
int nnzs;
|
||
|
|
};
|
||
|
|
|
||
|
|
void sortRow(int *colIndices, double *data, int left, int right);
|
||
|
|
|
||
|
|
} // end namespace bda
|
||
|
|
|
||
|
|
#endif // FPGA_MATRIX_HEADER_INCLUDED
|