A SCOOPP Evaluation on Packing Parallel Objects in Runtime.pdfVIP

A SCOOPP Evaluation on Packing Parallel Objects in Run-time* João Luís Sobral, Alberto José Proença Departamento de Informática - Universidade do Minho 4710 - 057 BRAGA – PORTUGAL {jls, aproenca}@di.uminho.pt Abstract The SCOOPP (Scalable Object Oriented Parallel Programming) system is an hybrid compile and run-time system. SCOOPP dynamically scales OO applications on a wide range of target platforms, including a novel feature to perform a run-time packing of excess parallel tasks. This communication details the methodology and policies to pack parallel objects into grains and method calls into messages. The SCOOPP evaluation focus on a pipelined parallel algorithm - the Eratosthenes sieve - which may dynamically generate a large number of fine-grained parallel tasks and messages. This case study shows how the parallelism grain-size - both computational and communication - has a strong impact on performance and on the programmer burden. The presented performance results show that the SCOOPP methodology is feasible and the proposed policies achieve efficient portability results across several target platforms. 1 Introduction Most parallel applications require parallelism granularity decisions: a larger number of fine parallel tasks may help to scale up the parallel application and it may improve the load balancing. However, if parallel tasks are too fine, performance may degrade due to parallelism overheads, both at the computational and communication level. Static granularity control, performed at compile-time, c