The Compute Unified Device Architecture (CUDA) enables NVIDIA
 graphics processing units (GPUs) to be used for massively parallel
 general purpose computation.
 .
 The cuSPARSE library contains a set of basic linear algebra subroutines used
 for handling sparse matrices. It is implemented on top of the NVIDIA CUDA
 runtime and is designed to be called from C and C++. The library routines can
 be classified into four categories:
  * Level 1: operations between a vector in sparse format and a vector in dense
 format
  * Level 2: operations between a matrix in sparse format and a vector in dense
 format
  * Level 3: operations between a matrix in sparse format and a set of vectors
 in dense format
  * Conversion: operations that allow conversion between different matrix
 formats
 .
 This package contains the cuSPARSE runtime library.