SHARCNET HPTC SUMMER SCHOOL 2009
June 1-5, 2009
Sheridan Institute of Technology and Advanced Learning
Trafalgar Campus, Oakville, Ontario



HOME
ABOUT
PROGRAMME
DAILY SCHEDULE
LOCATION & ACCOMMODATION
REGISTRATION
CALL FOR APPLICATION PROBLEMS
PARTICIPANTS
MAPS, TRAVEL INFO
CONTACT




SURVEY
GALLERY
PAST WORKSHOPS
SHARCNET

Optimization and Profiling

This course will cover basic concepts related to the optimization of large codes which are run frequently. It starts by describing how to obtain a profile (i.e. identifying which routines consume most of the execution time). Techniques to obtain a quick turn around for the code being optimized are presented (e.g. writing a prototype model). To illustrate these concepts a prototype model for a large climate model code (written in fortran) is introduced. Timing routines to measure segments of the code are used in the prototype model. The serial prototype model is timed and will be used as a reference for an MPI prototype model. As the MPI model is developed performance is measured against the serial version of the prototype model. Many short stand-alone programs will be presented to illustrate the optimization techniques. The selection of optimal number of processors will be illustrated through the MPI prototype model. The climate model uses a multigrid technique to solve the set of tridiagonal linear equations and it is written in fortran. All the material presented will be available in a tar file so that users will be able to refer to these programs.


© 2009 Shared Hierarchical Academic Research Computing Network
www.sharcnet.ca