Code optimization and performance tuning
This course provides an intermediate and advanced level treatment of all the essential ingredients for achieving high performance in numerical scientific computing on modern supercomputers. Computer architectures, compilers, libraries, as well as serial and parallel code optimization are explained and exemplified together with related hands-on exercises. The performance gains obtained by applying the described techniques can be very significant.
You are encouraged to bring your own piece of software for the hands-on sessions. If you wish the CSC experts to have a look on your software beforehand, please give the name and instructions for obtaining the source in the registration form.
Prerequisites
The participants are assumed to have working knowledge of either Fortran 90 or C programming languages as well as of parallel programming with MPI and/or OpenMP models.
Program
Mon 22 March
9.00-12.00 Introduction to performance analysis and optimization
12.00-13.00 Lunch
13.00-16.00 Optimal application porting
Tue 23 March
9.00-12.00 Single-core performance tweaking techniques
12.00-13.00 Lunch
13.00-16.00 Improving parallel scalability
Wed 24 March
9.00-12.00 Code optimization case studies
12.00-13.00 Lunch
13.00-16.00 Hands-on work
Each session consists of a lecture and hands-on exercise session, separated by a coffee break.
Registration
Additional information
Pekka Manninen, pekka.manninen@csc.fi, +358-9-457 2265
Sebastian von Alfthan, sebastian.von.alfthan@csc.fi, +358-9-457 2940
