Many operations in PARALLAX rely on sparse matrices, which are stored using the Compressed Sparse Row (CSR) format with one-based indexing. The csrmat_t type has the following attributes:
ndim: Number of rows.ncol: Number of columns.nnz: Number of non-zero elements.i: The row pointer (one-based indexing, dimension = ndim + 1).j: Column indices (one-based indexing, dimension = nnz).vals: Values of the matrix elements (dimension = nnz).There are also routines available to perform common sparse matrix-vector operations, as well as to sort (and sum duplicates of) entries in ascending order of column index for each row.