Performance Evaluation using the TAU Performance System
Registration (places are limited) and further information (to come):
http://agenda.albanova.se/conferenceDisplay.py?confId=4028
Abstract:
To meet the needs of computational scientists to evaluate the performance of their parallel, scientific applications, we present the TAU Performance System®[http://tau.uoregon.edu]. This one-day workshop will cover performance evaluation of applications on the Linux clusters with GPUs at PDC, KTH. This workshop will focus on instrumentation techniques provided by TAU and demonstrate how performance data (both profile and trace data) can be collected in a straightforward manner using TAU’s (Tuning and Analysis Utilities) automated instrumentation. It will also cover how to analyze the performance data collected and drill down to find performance bottlenecks and determine their causes. The workshop will include some sample codes that illustrate the different instrumentation and measurement choices available to the users. Topics will cover generating performance profiles and traces with heap memory usage, observed I/O bandwidth and volume, and hardware performance counters data using PAPI. Hardware counter data can show not only which routines are taking the most time, but why? For example, because of cache misses, TLB misses, excess address arithmetic, or poor branch prediction behavior. Users and developers are welcome to contact the instructor ahead of time to begin collecting data so as to have it on hand for the workshop. For more information see http://agenda.albanova.se/conferenceDisplay.py?confId=4028
