Bug Fixes  Fixed coo_matrix_view conversion error  Fixed interface error in LAPACK routines  Fixed memory access error in counting_sort  Fixed memory access error in DIA generalized_spmv  Fixed smoothed_aggregation and polynomial smoother errors related to VS2013  Added cusp/blas.h for compatibility with v0.4.0  Added convergence, default and verbose monitors for compatibility with v0.4.0  Added missing header files for size_t defs  Updated multiply interface for tighter integration with policy-based dispatching

Summary  Cusp v0.5.0 provides support for CUDA 7.0RC and Thrust 1.8. This release  reflects a major refactoring of the dispatch system and provides closer  integration with the internal structure of Thrust. This release also  marks a substantial improvement in the overall documentation of Cusp.New Features  Gauss-Seidel and SOR relaxation methods for CSR matrices  Implementation of robust Lanczos and LOBPCG eigensolvers (experimental)  Native implementation of BFS, imported from back40computing  Policy based interface to control execution and memory allocation of algorithms  Support for generalized SpMV operations sparse matrices in any format  Unified verbose_monitor and convergence_monitor functionality into the default monitor class  New IO routines to read graphs or matrices stored in DIMACS or binary format  counting_sort implementation to improve the performance of host sorting routines  join_iterator used to combine separate iterators in an arbitrary order  random_iterator used to generate random numbers  Algorithms     cusp::graph::vertex_coloring     cusp::relaxation::gauss_seidel     cusp::relaxation::sor     cusp::relaxation::jacobi     cusp::io::dimacs     cusp::io::binaryBug Fixes  Resolved issues related to Thrust 1.7 & Thrust 1.8  Resolved issues related to CUDA 6.5 & CUDA 7.0  Silenced warnings concerning Thrust OMP backend

Updated README

Verison 0.3.0

Verison 0.2.0

Verison 0.1.0

updated CHANGELOG with 0.3.1 info