Performance of ScaLAPACK

  

This chapter presents performance numbers for ScaLAPACK routines. The numbers are provided for illustration only and should not be regarded as a definitive up-to-date statement of performance. They have been selected from performance numbers obtained in 1996-1997 during the development of version 1.4 of ScaLAPACK. To obtain up-to-date performance figures, users should use the timing programs provided with ScaLAPACK.

 
• Achieving High Performance with ScaLAPACK
  • Achieving High Performance on a Distributed Memory Computer
  • Achieving High Performance on a Network of Workstations
• Performance, Portability and Scalability
  • The BLAS as the Key to (Trans)portable Efficiency
  • Two-Dimensional Block Cyclic Data Distribution as a Key to Load Balancing and Software Reuse
  • BLACS as an Efficient, Portable and Adequate Message-Passing Interface
    • Parallel Efficiency
  • ScaLAPACK Performance
  • Performance of Selected BLACS and Level 3 BLAS Routines
    • Performance of Selected PBLAS routines
    • Solution of Common Numerical Linear Algebra Problems
  • Solving Linear Systems of Equations
    • Solving Linear Least Squares Problems
    • Eigenvalue Problems
• Performance Evaluation
  • Obtaining High Performance with ScaLAPACK Codes
  • Checking the BLAS and BLACS Libraries
  • Estimate Execution Time
  • Determine Whether Reasonable Performance Is Achieved
  • Identify Performance Bottlenecks
  • Performance Bottlenecks in the Expert Symmetric Eigenproblem Drivers
• Performance Improvement
  • Choosing a Faster BLACS Library
  • Choosing a Faster BLAS Library
  • Tuning the Distribution Parameters for Better Performance
• Performance of Banded and Out-of-Core Drivers