High performance Fortran for highly irregular problems
PPOPP '97 Proceedings of the sixth ACM SIGPLAN symposium on Principles and practice of parallel programming
Processor reordering algorithms toward efficient GEN_BLOCK redistribution
Proceedings of the 2001 ACM symposium on Applied computing
Hi-index | 0.01 |
We have implemented the GEN_BLOCK (generalized block) data distribution in PGHPF, our High Performance Fortran implementation. Compared to a BLOCK or CYCLIC distribution, the more flexible GEN_BLOCK distribution allows users to balance the load between processors. Simple benchmark programs demonstrate the benefits of the new distribution format for unbalanced work loads, getting speedup of up to 2X over simple distributions. We also show performance results for a whole application.