Contents

• Overview and Motivation
• Design of ScaLAPACK
  • Portability, Scalability and Standards
  • ScaLAPACK Software Components
  • Processes versus Processors
  • Local Components
  • Block Cyclic Data Distribution
  • PBLAS
  • ScaLAPACK - LU Decomposition
  • ScaLAPACK - Symmetric Eigenproblem
  • Heterogeneous Networks
• The Distributed Linear Algebra Machine (DLAM)
  • The BLAS Process
  • The BLACS Network
  • Accuracy and Refinement of the DLAM
  • The LU factorization on the DLAM
• Performance
  • Choice of Block Size
  • Choice of Grid Size
• Future Directions
  • Future addition to ScaLAPACK
  • Alternative Approaches to Libraries
• References
• About this document ...