Performance Evaluation of A Seismic Data Analysis Kernel on The KSR MultiprocessorsReport as inadecuate


Performance Evaluation of A Seismic Data Analysis Kernel on The KSR Multiprocessors


Performance Evaluation of A Seismic Data Analysis Kernel on The KSR Multiprocessors - Download this document for free, or read online. Document in PDF available to download.

The paper investigates the effective performance attainable fora specific class of application programs on shared memory supercomputers.Specifically, we are to investigate how seismic data analysis applicationsbehave on the Kendall Square Research Inc.-s KSR multiprocessors. Thecomputational kernel of seismic computation algorithms is parallelized andits performance is analyzed. Three approaches for parallelizing the g5kernel are analyzed: column-based, row-based, and grid-basedparallelizations. All three approaches result in well balanceddecompositions, but differ significantly in data locality. In general, thecolumn-based approach has the best data locality, while the small grid-basedapproach has the worst. These results clearly indicate that data localityis one of the critical factors for attaining high performance for the g5kernel. The best parallelized g5 kernel code achieves about 44% of boththe KSR-1 and KSR-2 machines- peak computational performance.



College of Computing Technical Reports -



Author: Gu, Weiming - -

Source: https://smartech.gatech.edu/



DOWNLOAD PDF




Related documents