Exa-Tracer: Tracing HPC Supercomputers with LTTng

Over the past two years, the developers from EfficiOS have been working with the Argonne National Laboratory (THAPI), Lawrence Livermore National Laboratory, and AMD to adapt the LTTng tracer to meet the requirements from the HPC field.

The LTTng tracer provides flexible, low-overhead tracing of large multi-core systems. It has historically focused on the stringent requirements from the telecom and embedded industries, and is currently being adapted through the (Exa-Tracer) project to target HPC GPU-based supercomputers: instrumentation of MPI and ROCm APIs, scaling the tracer to analyze large clusters.