Abstract | ||
---|---|---|
Multiprocessors and high performance networks allow building CLUsters of MultiProcessors (CLUMPs). One dis- tinctive feature over traditional parallel computers is th eir hybrid memory model (message passing between the nodes and shared memory inside the nodes). We evaluate the performance of a cluster of dual processor PCs connected by a Myrinet network for NAS benchmarks using two pro- gramming models: a Single Memory Model based on the MPICH-PM/CLUMP library of the RWCP and a Hybrid Memory Model using MPICH-PM and OpenMP. MPI pro- grams are used as the reference in all experiments involv- ing programming models. We compare dual processor node configurations speedup versus uniprocessor node configu- rations for each model. We demonstrate that the superi- ority of one model over the other depends on the features of the applications. In particular, we detail the speedup results from breakdowns of the benchmark execution times and from measurements of hardware counters. |
Year | DOI | Venue |
---|---|---|
2000 | 10.1109/HPCA.2000.824364 | HPCA |
Keywords | Field | DocType |
shared memory,parallel computer,performance,memory model,programming models,programming model,read only memory,hidden markov models,message passing,hardware | Cluster (physics),Uniprocessor system,Shared memory,Programming paradigm,Computer science,Parallel computing,Real-time computing,Memory model,Message passing,Myrinet,Speedup | Conference |
Citations | PageRank | References |
16 | 1.99 | 12 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Franck Cappello | 1 | 3775 | 251.47 |
Olivier Richard | 2 | 16 | 1.99 |
Daniel Etiemble | 3 | 300 | 42.43 |