Speaker
Description
The talk will be given remotely and is fully confirmed to take place at the specified time.
Abstract:
In recent years, HPC systems have become increasingly complex and heterogeneous, including CPUs and GPUs, making application development and optimization challenging. Performance models like the Cache-aware Roofline Model (CARM) offer effective guidance by providing insights into bottlenecks that limit the application’s ability to reach the system’s maximum performance. The focus of this talk is the CARM Tool, which was developed to provide CARM support, not only to all major CPU architectures and ISAs, i.e., x86 (Intel, AMD), ARM, and RISC-V, but also to GPUs (NVIDIA). The CARM tool also covers specialized accelerators for artificial intelligence, such as the NVIDIA Tensor Cores, for the performance and efficiency characterization of deep learning workloads. The proposed tool includes automatically generated microbenchmarks, specifically tailored to cover both computational units and all memory hierarchy levels. Additionally, in this talk, I will showcase the deployment of CARM in the EuroHPC centers and present a short tutorial on the tool usage.
Speaker: Prof. Aleksandar Ilic
Aleksandar Ilic (PhD’14) is an Associate Professor at the Electrical and Computer Engineering Department (DEEC) of Instituto Superior Técnico (IST), Universidade de Lisboa, and a Senior Researcher of the INESC-ID, Lisbon, Portugal. His research interests include high-performance and energy-efficient computing and modeling of parallel heterogeneous systems. He has contributed to more than 90 international journal and conference publications.
