Performance Optimization and Modeling of Fine-Grained Irregular Communication in UPC

Table 2

Time usage (in seconds) of 1000 iterations SpMV for Test problem 1; naive UPC implementation (Listing 2) vs. the first transformed UPC implementation (Listing 3).

	1 thread	2 threads	4 threads	8 threads	16 threads
Naive UPC	895.44	548.57	301.17	173.08	106.10
UPCv1	270.40	159.51	86.37	51.10	28.80