Obtaining High Performance

• Use the best BLAS and BLACS libraries available.

• Start with a standard data distribution.

  • A square processor grid (Pr=Pc) if P >= 9
  • A one-dimensional processor grid (Pr=1,Pc=P) if P < 9
  • Block size (NB) = 64

• Determine whether reasonable performance is being achieved.

• Identify the performance bottlenecks, if any.

Previous slide

Next slide

Back to first slide

View graphic version