Basic HTML version of Foils prepared 15 March 1996

Foil 4 General Analysis of Parallel ADI Performance -- Dependence of Computation and Communication

From Further PDE Solvers for the NAS Benchmarks CPSP713 Case studies in Computational Science -- Spring Semester 1996. by Geoffrey C. Fox
Just consider dependence on NNAS and number of Processors P
  • Ignore constants
Pipeline Start-up latency is proportional to NNAS
Total Computation 3N 3NAS/P
Communication:
  • 3 Dimensional Decomposition: 3N 2NAS / P2/3
  • 1 Dimensional Decomposition: N 2NAS
Alternative appreoach is Redistribution to arrange that ADI always inside node
  • Redistribution Time proportional to N 3NAS / P
  • No Pipeline Start-Up

Northeast Parallel Architectures Center, Syracuse University, npac@npac.syr.edu

If you have any comments about this server, send e-mail to webmaster@npac.syr.edu.

Page produced by wwwfoil on Sun Feb 22 1998