Vierailijaluento: José R. Herrero "Profiling parallel applications with Paraver"

Profiling parallel applications with Paraver

José R. Herrero
Computer Architecture Department
Universitat Politècnica de Catalunya - BarcelonaTech

Abstract: Writing parallel applications that make good use of the resources is not an easy task. Performance analysis tools support developers on the evaluation, tuning and optimization of their codes. In this seminar we will present some performance tools developed at Barcelona Supercomputing Center (BSC). They are an open-source project targeting not only to detect performance problems but to understand the applications' behavior.

The key component is Paraver, a performance analyzer based on traces with a great flexibility to explore the collected data. Paraver was developed to respond to the need to have a qualitative global perception of the application behavior by visual inspection and then to be able to focus on the detailed quantitative analysis of the problems. 

Extrae is the package devoted to generating Paraver trace-files for a post-mortem analysis. Extrae is a tool that uses different interposition mechanisms to inject probes into the target application so as to gather information regarding the application performance. Currently it supports several programming interfaces such as MPI, OpenMP, OpenCL, pthreads, OmpSs, Java and CUDA.

Paraver is not tied to any programming model as long as the model used can be mapped in the three levels of parallelism expressed in the Paraver trace. An example of a two-level parallelism would be that of hybrid MPI + OpenMP applications.

In this seminar we will try to illustrate how these tools can be used to analyze the performance of parallel applications and to familiarize the audience with the usage of these tools.