Pilgrim - A Lossless MPI Tracing Tool

Chen Wang, University of Illinois Urbana-Champaign

Abstract: In this talk, we will describe a new MPI tracing tool, Pilgrim. Unlike existing tools that only keep summarized information, Pilgrim is lossless in that it records all parameters of every MPI function call. To achieve this, the size of the generated trace files should remain at an acceptable level regardless the problem size and the scale of runs. We will be discussing the encoding and compression mechanisms that enable this "lossless" feature. We will also present the results from several scientific applications.

Bio: Chen Wang, a 4th year CS PhD student at University of Illinois at Urbana-Champaign. His advisor is Prof. Marc Snir. His research interest mainly focuses on high-performance computing and parallel storage systems.

Please use this link to attend the virtual seminar: