Sometimes memory has become a major problem in applications, with its bandwidth but also by the incresing size needed by more and more complex and dynamic applications. So, how to track these errors and point problematic patterns ? How to find where the memory is consumed when the application reaches the hardware limit ? After my PhD on memory management in HPC context (NUMA, parallel, etc) I had the opportunity to develop two profilers (malloc and numa) now open-sources for C/C++/Fortran and Rust. I will briefly present these tools with some examples and expected observations.
What are Gray Scott Thursdays?
Gray Scott Thursdays are a unique series of 18 live webinars, held every Thursday leading up to the Gray Scott School 2026. Each session is presented by our expert professors and provides an early dive into the key topics of High-Performance Computing, including:
CPU/GPU architectures, memory management, profiling, testing, and modern programming languages such as C++, Fortran, Julia and Python… and much more!
These webinars are a perfect opportunity to get a head start, learn directly from the instructors, and familiarize yourself with the tools, libraries, and methods that will be used during the intensive on-site training.
