Sparse matrix-vector multiplication

This repo is a basic implementation of multiplication between a sparse matrix and a vector and its parallelization using CUDA C. In csrmatrix_product.cu there is an example which has been tested on google colab with a matrix of NxN dimension with N = 40000 and a block dimension of 1024 which gained a result in 0.000027s. It is present also an implementation of the CSR Matrix struct in C to represent a sparse matrix, with methods to calculate the product with a vector that, with N=40000, cannot be compared with the Cuda results as the struct of the matrix in C represent a bottleneck for the cpu procedure.

LAST UPDATE

main.cu compares csr matrix multiplication on cpu and the cuda product on a matrix 51813503x51813503 with 103565681 values.

RESULT EXEMPLE

$ ./main 1024 DATA_PATH
CPU EXEC TIME: 1.149505s
GPU EXEC TIME: 0.000073s