Sparse Matrix Multiple Vector Multiplication using Ellpack storage format (SpMM_ELL)

Sparse matrix multi vector multiplication

This project reusues code from Nvidia's open source CUSP Library.

Related publication available here: http://ieeexplore.ieee.org/abstract/document/7056883/?reload=true

Compilation Command:

nvcc ELL.cu mmio.c -o ell_SpMM

Sample Output using CUDA/9.1 on V100 GPU:

Using 64-bit floating point precision

Reading sparse matrix from file (/scratch/cant.mtx): done
Using 62451-by-62451 matrix with 4007383 nonzero values
###   Testing the performance of SpMM using ELL   ###
Number of vectors 2    
	benchmarking ell                  [gpu]:   0.0791 ms ( 202.63 GFLOP/s)
	benchmarking ell                  [gpu]: ( 1113.24 Gbytes/s)
###   Testing the performance of SpMM using ELL   ###
Number of dense vectors 4    
	benchmarking ell                  [gpu]:   0.1056 ms ( 303.60 GFLOP/s)
	benchmarking ell                  [gpu]: ( 833.99 Gbytes/s)
###   Testing the performance of SpMM using ELL   ###
Number of dense vectors 8    
	benchmarking ell                  [gpu]:   0.1498 ms ( 427.89 GFLOP/s)
	benchmarking ell                  [gpu]: ( 587.70 Gbytes/s)
###   Testing the performance of SpMM using ELL   ###
Number of dense vectors 16    
	benchmarking ell                  [gpu]:   0.2959 ms ( 433.43 GFLOP/s)
	benchmarking ell                  [gpu]: ( 297.66 Gbytes/s)
###   Testing the performance of SpMM using ELL   ###
Number of dense vectors 32    
	benchmarking ell                  [gpu]:   0.6233 ms ( 411.45 GFLOP/s)
	benchmarking ell                  [gpu]: ( 141.28 Gbytes/s)