THE WORKSHOP
Heterogeneity is emerging as one of the main characteristics of today’s and future HPC environments where different nodes organizations, memory hierarchies, and kinds of exotic accelerators are increasingly present. It pervades the entire spectrum of Computing Continuum, ranging from large Cloud infrastructures and Datacenter up to the Internet of Things and Edge Computing environments, aimed at making available in a transparent and friendly way the multitude of heterogeneous HPC resources available everywhere around us.
In this context, for Computational Science and Machine Learning, it is essential to leverage efficient and highly scalable libraries and tools capable of exploiting such modern heterogeneous computers. These systems are typically characterized by very different software environments, which require a new level of flexibility in the algorithms and methods used to achieve an adequate level of performance, with growing attention to energy consumption.
This workshop aims to provide a forum for researchers and practitioners to discuss recent advances in parallel methods and algorithms and their implementations on current and future heterogeneous HPC environments. We solicit research works that address algorithmic design, implementation techniques, performance analysis, integration of parallel numerical methods in science and engineering applications, energy-aware techniques, and theoretical models that efficiently solve problems on heterogeneous platforms.
TOPICS
We focus on papers covering various topics related to the development of scientific applications on heterogeneous HPC environments with special interest to Computational Science and Machine Learning, that include, but are not limited to the following:
- Scalable algorithms for heterogeneous environments
- Tools and programming environments supporting different forms of parallelism
- Multi/Many-cores and GPU tools for large-scale problems
- Innovative HPC computing devices
- Federated learning techniques
- Performance and scalability models
- Energy aware algorithms on low power devices
- Adaptive techniques for heterogeneous environments
- Analysis methods for large and non-structured data sets
- Memory management on non-uniform memory access environments
- Task scheduling and load balancing among heterogenous computing elements
- Integration of Cloud/Fog/Edge computing techniques and tools
- Scalable and efficient scientific workflows orchestration
ORGANIZERS
- Salvatore Cuomo (Univ. of Naples Federico II, Italy)
- Marco Lapegna (Univ. on Naples Federico II, Italy )
- Francesco Piccialli (Univ. of Naples Federico II)
- Diego Romano (Italian National Research Council)
PROGRAM:
The workshop will be held on wednesday 12 march at 11.30.
- The P3 Explorer: An Open Database of Performance, Portability, and Productivity – Steven Wright (University of York – UK)
- On the Effectiveness of Unified Memory in Multi-GPU Collective Communication – Ian Di Dio Lavore (Politecnico di Milano – IT)
- G-Litter Marine Litter Dataset Augmentation with Diffusion Models and Large Language Models on GPU Acceleration – Gennaro Mellone (University of Naples Parthenope – IT)
- NAV: A Comparative Analysis Tool for Nsight Systems GPU Traces – Ethan Shama (Queen’s University – CA)
- A Multi-Level Parallel Algorithm for Detection of Single Scatterers in SAR Tomography – Massimiliano Russo (University of Naples Federico II – IT)