A Novel CSR-Based Sparse Matrix-Vector Multiplication on GPUs

Algorithm 2

Kernel 1.

Input:

   CUDA-specific variables:

   (i) threadId.x: a thread

   (ii) blockId.x: a block

   (iii) blockDim.x: number of threads per block

   (iv) gridDim.x: number of blocks per grid

Output:

(01) threadId.x + blockId.x blockDim.x;

(02) blockDim.x gridDim.x;

(03) while

(04)   ;

(05)   += ;

(06) end while