Performance Tuning and OpenMP

14.01.02


Start Lecture by clicking here


Content

Author: Matthias Mueller

  1. Performance Tuning and OpenMP
  2.     Outline
  3.     Motivation
  4.     Performance Basics: Speed Up
  5.     Performance Basics: Amdahl’s Law
  6.     Consequence of Amdahl’s law: necessary parallelization
  7.     Performance measurement: Profiling
  8.     General issues: Problem size dependency of performance
  9.     OpenMP version of Matrix Norm
  10.     Performance on a PC (Dual Pentium II, 450 MHz)
  11.     Performance on a Vectorcomputer (NEC SX-5Be)
  12.     Performance on the Hitachi SR8000
  13.     Performance comparison (2 threads with OpenMP)
  14.     Use OpenMP only with sufficient workload: if-clause
  15.     Performance with if-clause
  16.     Performance with OpenMP: Avoid thread creation
  17.     Performance with OpenMP: Avoid barriers
  18.     Performance with OpenMP: Avoid barriers (II)
  19.     Performance with OpenMP: load balancing
  20.     Summary
  21.     Exercise: Matrix Norm Calculation
  22.     Exercise: Matrix Norm Calculation on SR8000
  23.     Exercise: Matrix Norm Calculation on SR8000
  24.     Exercise: Matrix Norm Calculation on SR8000
  25.     Exercise: Matrix Norm Calculation
  26.     Gnuplot

Back to the Parallel Programming Workshop Overview