NECSTFridayTalk – Exploring Architectural-Aware Affinity Policies in Modern HPC Runtimes
NECSTFridayTalk
Speaker: Ian Di Dio Lavore
PHD Student in Information Technology
DEIB - NECSTLab Meeting Room (Bld. 20)
Online by Zoom
October 4h, 2024 | 11.30 am
Contact: Prof. Marco Santambrogio
Speaker: Ian Di Dio Lavore
PHD Student in Information Technology
DEIB - NECSTLab Meeting Room (Bld. 20)
Online by Zoom
October 4h, 2024 | 11.30 am
Contact: Prof. Marco Santambrogio
Abstract
On October 4th, 2024 at 11.30 am a new appointment of NECSTFridayTalk series titled "Exploring Architectural-Aware Affinity Policies in Modern HPC Runtimes" will take place at DEIB NECSTLab Meeting Room (Building 20) and on line by Zoom.
During this talk, we will have, as speaker, Ian Di Dio Lavore, PhD student at Dipartimento di Elettronica, Informazione e Bioingegneria.
Modern commodity and High-Performance Computing (HPC) systems feature increasingly complex CPU architectures with higher core counts, NUMA domains, and hyperthreading. These varying hardware configurations, software stacks, and libraries make performance portability challenging, placing a significant burden on developers to optimize applications for each architecture. This work examines how hardware-dependent factors, such as locality and thread affinity, impact performance in modern CPUs. Using the Global Memory and Threading (GMT) runtime system, we evaluate existing and new affinity policies, explore NUMA configurations, and conduct scalability studies on three HPC clusters. Our findings highlight the critical role of proper runtime configurations in optimizing performance on complex architectures.
During this talk, we will have, as speaker, Ian Di Dio Lavore, PhD student at Dipartimento di Elettronica, Informazione e Bioingegneria.
Modern commodity and High-Performance Computing (HPC) systems feature increasingly complex CPU architectures with higher core counts, NUMA domains, and hyperthreading. These varying hardware configurations, software stacks, and libraries make performance portability challenging, placing a significant burden on developers to optimize applications for each architecture. This work examines how hardware-dependent factors, such as locality and thread affinity, impact performance in modern CPUs. Using the Global Memory and Threading (GMT) runtime system, we evaluate existing and new affinity policies, explore NUMA configurations, and conduct scalability studies on three HPC clusters. Our findings highlight the critical role of proper runtime configurations in optimizing performance on complex architectures.
The NECSTLab is a DEIB laboratory, with different research lines on advanced topics in computing systems: from architectural characteristics, to hardware-software codesign methodologies, to security and dependability issues of complex system architectures.
Every week, the “NECSTFridayTalk” invites researchers, professionals or entrepreneurs to share their work experiences and projects they are implementing in the “Computing Systems”.