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

Table 6

Comparison of PCSRI and PCSRII with communication on four GPUs.

Matrix	ET (GPU)	PCSRI (4 GPUs)			PCSRII (4 GPUs)
		ET	SD	PE	ET	SD	PE
2cubes_sphere	0.4444	0.1567	0.0052	70.89	0.1531	0.0028	72.54
scircuit	0.3484	0.1544	0.0204	56.39	0.1495	0.0073	58.27
Ga41As41H72	4.2387	1.7157	0.7909	61.76	1.4154	0.2178	74.87
F1	6.5544	2.1149	0.3833	77.48	2.0022	0.1941	81.84
ASIC_680ks	0.8196	0.3449	0.0187	59.39	0.3423	0.0147	59.87
ecology2	1.2321	0.4257	0.0048	72.35	0.4257	0.0056	72.35
Hamrle3	1.7684	0.6231	0.0087	70.95	0.6297	0.0085	70.21
thermal2	2.0708	0.6922	0.0267	74.78	0.6959	0.0269	74.39
cage14	5.9177	1.9339	0.3442	76.50	1.6417	0.0067	90.12
Transport	4.7305	1.3323	0.0279	88.77	1.3217	0.0070	89.48
G3_circuit	1.9731	0.7234	0.0408	68.19	0.7458	0.0620	66.14
kkt_power	4.3465	1.7277	0.5495	62.89	1.3791	0.0305	78.79
CurlCurl_4	5.1605	1.5065	0.0253	85.63	1.5004	0.8789	85.99
memchip	3.8257	1.3804	0.1768	69.29	1.3051	0.1029	73.28
Freescale1	5.0524	2.0711	0.4342	60.98	1.8193	0.2262	69.43