Extrae.jl: Advanced Profiling of Julia Code in HPC Clusters

Explore advanced profiling techniques for Julia code in High-Performance Computing (HPC) clusters using Extrae.jl. Dive into the capabilities of this powerful tool, which extends beyond Julia's native profiling methods. Learn how to annotate user regions, sample hardware counters, inspect callstacks inside C libraries, mark inter-node, inter-process, and inter-thread communication, intercept MPI, CUDA, and OpenMP calls, and emit custom user events. Witness practical demonstrations of performance evaluation for scientific applications written in Julia on x86_64 and AArch64 architectures, including those with scalable vector ISA and unified memory between CPU and GPU. Gain insights into understanding and optimizing performance behavior in complex HPC environments through real-world examples and expert guidance.

Extrae.jl: Advance profiling of Julia code in HPC clusters | Giordano, Sánchez Ramírez, Garcia Saez