Release Notes#

This section includes significant changes, new features, performance improvements and various issues.

NVPL SPARSE 0.4.1#

Added nvpl_sparse_const_dn_vec_get and nvpl_sparse_const_dn_vec_get_values dense vector API
Added nvpl_sparse_const_sp_mat_get_values to general sparse matrix API
Added nvpl_sparse_csr_get, nvpl_sparse_const_csr_get, and nvpl_sparse_csr_set_pointers to CSR matrix API
Added nvpl_sparse_csc_get, nvpl_sparse_const_csc_get, and nvpl_sparse_csc_set_pointers to CSC matrix API
Added nvpl_sparse_coo_get, nvpl_sparse_const_coo_get, and nvpl_sparse_coo_set_pointers to COO matrix API

Added nvpl_sparse_get_version API to obtain the library version
Added new dense vector utility routines: nvpl_sparse_dn_vec_get, nvpl_sparse_dn_vec_get_values, and nvpl_sparse_dn_vec_set_values
Added nvpl_sparse_spsv_update_matrix API to update sparse matrix values between analysis and solve routines

The first early access release of NVPL SPARSE library.

Sparse C APIs support
Supports single- and multi-threaded (based on OpenMP) computation
Supports different sparse matrix data layouts and storage formats, such as CSR, COO, CSC, and Sliced-ELL
Supports different types of matrix/vector indices (e.g., 32-bit, 64-bit)
Supports different input/output/compute data types
Added generic sparse matrix-vector (SpMV) API
Added generic sparse triangular solver (SpSV) API
Added extensive consistency checks across input matrices and vectors. This includes the validation of sizes, data types, layout, allowed operations, etc
Provides constant descriptors for vector and matrix inputs to support const-safe interface and guarantee that the APIs do not modify their inputs