Privacy & Legal Notice

Livermore Computing Training Announcement

Title: Parallel Performance Analysis with Open|SpeedShop (O|SS) & CBTF
Date/Times
Locations:
Aug 21, 2014
1:00pm-3:00pm: B451 R1025 (White Room) - Presentation & Demos
3:00pm-5:00pm: B453 R1016 (Viz Lab) - Individual sessions scheduled through Martin Schulz (schulzm@llnl.gov)
Aug 22, 2014
9:00am-5:00pm: B453 R1016 (Viz Lab) - Individual sessions scheduled through Martin Schulz (schulzm@llnl.gov)
Presentors: Jim Galarowicz and Don Maghrak from the Krell Institute
Description: On August 21 and 22, 2014, the two main developers of O|SS and the Component Based Tool Framework (CBTF), Jim Galarowicz and Don Maghrak from the Krell Institute, will present an overview of these toolsets (both basic functionality and new features) to demonstrate their features and provide direct support with your use of O|SS with LLNL codes. They will also be available to meet with interested code developers in one-to-one sessions. If you are not yet familiar with O|SS, this is a tremendous opportunity to get introductory instructions. Furthermore, if you are already familiar with O|SS, but desire to learn the new features of O|SS or have specific problems addressing performance bottlenecks in your codes, or if you want to give O|SS a try with your code, please join us for one of the 1:1 hands-on sessions. If you are interested in learning the new features of CBTF that uses state-of-the-art dynamic instrumentation, MRnet and filters for intelligent data reduction for large parallel runs for your own data analysis/reduction needs this will be an excellent opportunity to share your ideas and make requests on features you would like to see developed. Please R.S.V.P if you would like to attend a 1:1 sessions by emailing Martin Schulz (schulzm@llnl.gov)

Open|SpeedShop (O|SS) and CBTF are open source performance analysis toolsets targeting sequential codes, large scale parallel applications based on MPI, and multithreaded codes. They provide a large range of performance experiments covering most typical performance analysis needs including statistical runtime sampling, hardware counter analysis, MPI and I/O profiling, as well as floating point exception detection. O|SS specifically targets both software developers and performance analysis experts offering multiple interfaces including a comprehensive graphical user interface with performance analysis wizards, a textual command line interface that allows batch processing, and a Python interface for scripting support. These toolsets are being developed and maintained in a partnership between the Krell Institute and ASC/NNSA. It is currently installed and maintained on LLNL's TLCC machines (CZ and RZ) and BG/Q machines.

Information on O|SS can be found at: http://www.openspeedshop.org/wp/

Fee: None
Registration: No registration is required. Seating is on a first-come, first-served basis.
Questions? Please contact Martin Schulz (schulzm@llnl.gov)