Add CUDA-Enabled Matrix Multiplication

PASSIONLab / CombBLAS

The Combinatorial BLAS (CombBLAS) is an extensible distributed-memory parallel graph library offering a small but powerful set of linear algebra primitives specifically targeting graph analytics.

Other

64 stars 22 forks source link

Add CUDA-Enabled Matrix Multiplication #22

Open tf-mac opened 7 months ago

tf-mac commented 7 months ago

This adds a couple of changes to this project

Enables CUDA in the project, under the __CUDACC__ define
Adds a CUDA-aware MPI distribution scheme
Adds GALATIC to the includes, allowing access to its methods
Combines all of these to add the method Mult_AnXBn_DoubleBuff_CUDA

Initial results are below: delaunay matrix runtime (s), 183M (n=100) rmat matrix runtime (s), 8M (n=100)

richardlett commented 6 months ago