Binary Matrix (BMT) file format

BMT is a binary (as opposed to ASCII) format for storing matrices. It supports

• Sparse and dense matrices.
• Row major (resp. CSR) and column major (resp. CSC) matrices.
• Double or single precision entries.
• 32- or 64-bit integer indices.

A MATLAB interface is provided, making it easy to export matrices from MATLAB to C++ and vice versa (no more need to deal with pesky MatrixMarket files). The format is intended to be exceedingly simple and as a result does not support compression or multiple matrices in one binary.

Examples

Examples are provided for both MATLAB and C++. For MATLAB specific examples see matlab/test_bmt.m. This is a script, which can be executed as is. For C++ see test/test_bmt.cpp, to run the test, simply type make test.

We provide two examples of exporting matrices in MATLAB and loading them in C++.

Sparse matrix example

To export a sparse matrix from MATLAB, type

 bmtsave(A, 'A_sparse.bmt')

Then to load it in C++

char order;
char fname[] = "A_sparse.bmt";
int m, n, nnz;
double *val;
int  *col_ind, *row_ptr;
ReadBmtSparse(fname, &order, &m, &n, &nnz, &val, &col_ind, &row_ptr);

Upon exiting, variables take the following values

• order: The character 'c', indicating that the matrix was stored in column major ordering.
• m, n: Dimensions of matrix A.
• nnz: Number of non-zero entries in A.
• col_ind: Indicies of of columns for each element.
• row_ptr: Pointer to beginning of each rows.

Dense matrix example

The syntax in MATLAB is the same for sparse and dense matrices:

 bmtsave(A, 'A_dense.bmt')

Then, assuming A is dense, we can load it in C++

char order;
char fname[] = "A_dense.bmt";
int m, n;
double *A;
ReadBmtDense(fname, &order, &m, &n, &A);

Upon exiting, variables take the following values

• order: The character 'c', indicating that the matrix was stored in column major ordering.
• m, n: Dimensions of matrix A.
• A: Entries in A.