SC20 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Enabling Graph-Based Profiling Analysis Using Hatchet


Student: Ian D. Lumsden (University of Tennessee)
Supervisor: Michela Taufer (University of Tennessee, Knoxville)

Abstract: We augment the Hatchet performance data analysis tool to leverage relational caller-callee data in the analysis of HPC applications. Specifically, we design a query language that enables data reduction using graph path pattern matching. We demonstrate the capabilities of our augmented Hatchet by examining the performance of MPI calls in three High Performance Computing (HPC) benchmarks when using MVAPICH and Spectrum-MPI. In doing so, we can identify hidden performance losses in specific MPI functions.

ACM-SRC Semi-Finalist: yes

Poster: PDF
Poster Summary: PDF


Back to Poster Archive Listing