Research Article
Efficient CSR-Based Sparse Matrix-Vector Multiplication on GPU
Table 4
Speedups of IPCSR over other implementation techniques in double precision.
| Matrix | Speedups on C2050 | Speedups on K20c | CUSPARSE | CUSP | clSpMV | PCSR | CSR5 | CSR-Adaptive | CUSPARSE | CUSP | clSpMV | PCSR | CSR5 | CSR-Adaptive |
| Dense | 0.91 | 0.95 | 2.06 | 2.72 | 1.01 | 1.00 | 1.08 | 0.97 | 2.01 | 2.68 | 1.05 | 1.01 | Protein | 0.87 | 0.92 | 0.94 | 1.08 | 1.01 | 0.95 | 1.09 | 1.10 | 0.94 | 1.09 | 1.08 | 0.99 | FEM/Spheres | 1.37 | 1.39 | 1.29 | 1.36 | 1.22 | 1.00 | 1.72 | 1.66 | 1.29 | 1.37 | 1.30 | 1.07 | FEM/Cantilever | 1.35 | 1.41 | 1.26 | 1.55 | 1.17 | 1.04 | 1.70 | 1.77 | 1.26 | 1.57 | 1.25 | 1.11 | Wind Tunnel | 1.34 | 1.21 | 0.82 | 1.27 | 0.86 | 0.83 | 1.82 | 1.69 | 0.88 | 1.38 | 1.00 | 0.97 | FEM/Harbor | 1.13 | 1.22 | 1.05 | 1.28 | 1.06 | 1.02 | 1.42 | 1.35 | 1.05 | 1.29 | 1.14 | 1.09 | QCD | 1.49 | 0.71 | 0.96 | 1.16 | 1.05 | 1.01 | 1.87 | 0.77 | 0.70 | 1.17 | 1.13 | 1.09 | FEM/Ship | 1.43 | 1.60 | 1.19 | 1.23 | 1.08 | 1.01 | 1.79 | 1.89 | 1.19 | 1.24 | 1.15 | 1.08 | Economics | 2.28 | 1.68 | 1.67 | 1.59 | 1.15 | 1.056 | 2.87 | 2.04 | 1.67 | 1.60 | 1.23 | 1.08 | Epidemiology | 1.42 | 0.90 | 0.74 | 1.07 | 1.01 | 1.00 | 1.79 | 1.10 | 0.74 | 1.08 | 1.08 | 1.03 | FEM/Accelerator | 1.70 | 1.65 | 0.92 | 1.48 | 1.07 | 1.00 | 2.13 | 1.37 | 0.92 | 1.49 | 1.15 | 1.07 | Circuit | 1.94 | 1.85 | 1.25 | 1.11 | 1.06 | 1.02 | 2.56 | 2.36 | 1.31 | 1.17 | 1.19 | 1.06 | Webbase | 6.11 | 1.29 | 0.93 | 1.19 | 1.17 | 1.05 | 7.93 | 1.63 | 0.96 | 1.24 | 1.29 | 1.08 | LP | 1.39 | 1.28 | 1.29 | 1.79 | 1.28 | 1.01 | 1.75 | 1.64 | 1.18 | 1.81 | 1.37 | 1.08 | circuit5M | 7.61 | 1.71 | 1.15 | 1.74 | 1.10 | 1.06 | 9.80 | 1.64 | 1.18 | 1.80 | 1.21 | 1.16 | eu-2005 | 1.90 | 1.78 | 1.52 | 2.06 | 1.17 | 1.03 | 2.44 | 2.22 | 1.56 | 2.13 | 1.29 | 1.13 | Ga41As41H72 | 1.06 | 1.19 | 1.09 | 1.23 | 1.08 | 1.03 | 1.33 | 1.70 | 1.09 | 1.24 | 1.16 | 1.11 | in-2004 | 1.69 | 1.41 | 1.31 | 1.92 | 1.23 | 1.14 | 2.18 | 2.08 | 1.35 | 1.98 | 1.35 | 1.26 | mip1 | 1.10 | 1.11 | 1.16 | 1.87 | 1.08 | 0.98 | 1.39 | 1.59 | 1.16 | 1.88 | 1.15 | 1.03 | Si41Ge41H72 | 1.45 | 1.29 | 1.97 | 2.50 | 1.26 | 1.01 | 1.82 | 1.84 | 1.80 | 1.99 | 1.35 | 1.08 |
|
|