Improving Application Performance Using TAU

Date and Time: 
2013 Wednesday, April 3
CG1 Auditoriums
John Linford

Authors: John Linford and Sameer Shende

To meet the needs of computational scientists to evaluate and improve the performance of their parallel, scientific applications, we present the TAU Performance System. This workshop will focus on performance data collection, analysis, and performance optimization. The tutorial will introduce profiling and debugging support in TAU. TAU now includes support for tracking callstacks at the point of program failure to isolate runtime faults. The tutorial will cover performance evaluation of parallel programs written in Python, Fortran, C++, C, and UPC, using MPI, and other runtime layers such as CUDA, OpenCL, SHMEM, and OpenMP with the TAU Performance System. In this tutorial, we will demonstrate different techniques for program instrumentation. The talk will also highlight TAU's support for memory debugging, and I/O evaluation. The hands-on portion of the tutorial will guide the developers through the instrumentation, measurement, and analysis process steps int TAU. We will introduce new instrumentation techniques to simplify the usage of performance tools. These will include compiler-based instrumentation, binary re-writing, library preloading for CUDA instrumentation, native and MIC offloading based modes for Intel Xeon Phi coprocessors, and automatic instrumentation of source code. Performance data will include MPI timings, GPGPU transfers, runtime bounds checking, I/O and memory, and hardware performance counters from PAPI. The tutorial will demonstrate how TAU's instrumentation and analysis tools may be used with external tools such as Score-P, Scalasca, OTF2, PAPI, and Vampir. For the hands-on session, the participants will be able to use NCAR systems or an optional HPC Linux LiveDVD that will allow them to boot their laptops to a Linux distribution that has the above tools installed. The participants are encouraged to bring a laptop with them and install VirtualBox virtualization software and related OVA files from

Speaker Description: 

Dr. John Linford is a Scientist at ParaTools, Inc. He received his Ph.D. from Virginia Tech, where his dissertation on accelerating atmospheric modeling through emerging multi-core technologies was selected as the outstanding doctoral dissertation of 2010. John has developed a meta-programmer for chemical kinetic simulation, airborne signal processing applications, rotocraft engineering tools, and toolkits for porting parallel HPC applications to cloud computing platforms. John helps develop the TAU Performance System and has contributed to the Scalasca project and the MoinMoin project.

PDF icon tau_sea13_15min.pdf3.9 MB
Video recorded: 

Event Category: