Exploiting Delayed Synchronization Arrivals in Light-Weight Data Parallelism

SPMD Single Program Multiple Data models and other traditional models of data parallelism provide parallelism at the processor level Barrier synchroniza tion is de ned at the level of processors where when a processor arrives at the barrier point early and waits for others to arrive no other useful work is done on that processor Program restructuring is one way of min imizing such latencies However such programs tend to be error prone and less portable In this paper we discuss how multithreading can be used in data paral lelism to mask delays due to application irregularity or processor load imbalance The discussion is in the con text of Coir our object oriented runtime system for parallelism The discussion concentrates on shared memory systems The sample application is an LU factorization algorithm for skyline sparse matrices We discuss p erformance results on the IBM PowerPC based symmetric multiprocessor system