As machines get larger and scientific applications advance, it is more and more imperative to fully utilize high performance computing (HPC) capability. The complexity and changing landscape of parallel computers may lead to users being unable or unsure how to achieve optimal performance from their applications. This lecture covers the performance analysis methodology of the POP project, aiming to improve the efficiency of compute-demanding codes and export a set of best practices for programmability to the HPC community. The attendants will get familiar with our general-to-detail analysis methodology, that goes through different steps in order to first detect potential scalability bottlenecks, and then go deep into the causes of these problems and the specific code regions affected. Two key elements to achieve this goal will be the tools used, and the set of efficiency metrics that enable the analyst to get a first insight of the application’s performance.
Speaker
Marc Clascà Ramírez – Barcelona Supercomputing Center
Marc Clascà is a master student in Computer Engineering at Universitat Politècnica de Catalunya, and a research engineer at the Barcelona Supercomputing Center since 2020. His research interests include programming models, performance tools, performance analysis and specialized hardware and accelerators. He works in HPC parallel performance analysis in the Best Practices for Performance and Programmability (BePPP) group, aiming to provide scientific developers with the best practices in programming portable and performant codes. His current research focus is analyzing the performance of scientific applications and AI use cases that use accelerators to extract new metrics and methodologies. This includes exploring the potentials of GPU specific tracing and visualization tools, and understanding new programming models and communication patterns used in LLM training and inference.
Event Timeslots (1)
Fri 20 – Applications & Performance Evaluation
-
M. Clascà Ramírez (Barcelona Supercomputing Center)