Efficient CSR-Based Sparse Matrix-Vector Multiplication on GPU

Figure 7

Effective bandwidth of all algorithms on C2050.