Efficient CSR-Based Sparse Matrix-Vector Multiplication on GPU

Figure 4

Average speedups of IPCSR over other implementation techniques.