Sparse Matrix-Vector Multiplication on GPGPUs