Review of Techniques I

Review of Techniques I

• Select algorithm for effective parallelism
  • Don�t sacrifice elapsed-time performance for artificial �efficiency� measurements
• Choose effective data decompositions
  • Tune to problem sizes of interest
  • Look carefully at parts of code that don�t scale with p
• Model and measure
  • Look for underperforming sections of code

Previous slide

Next slide

Back to the first slide

View Graphic Version