The 1st Workshop on Heterogeneous Memory Systems (HMEM 2020), collocated with ICS 2020, was held virtually on 29 June 2020. It was attended by 38 people, who were given the opportunity to hear about ongoing research around heterogeneous memory systems and discuss the topic with the panellists.
Talk 1: Libmemkind - Heap Manager for heterogeneous memory architectures by Michal Biesek (Intel)
Michal gave an overview about the current status and possibilities of the memkind library, which is able to handle and manage combinations of diverse memory technologies.
Talk 2: An Ecosystem of Tools for Broad Heterogeneous Memory Usage by Marc Jordà (BSC)
Marc introduced the software ecosystem developed under the Intel-BSC Exascale Laboratory umbrella and in collaboration with the EPEEC project. The aim of that software is to facilitate the efficient use of heterogeneous memory systems, and its early performance analysis.
Talk 3: Abstractions for Data Management in Heterogeneous Systems by Tim Dykes (Hewlett Packard Enterprise, HPE)
Tim presented the work of the EPiGRAM-HS project on Mamba library, an array-based abstraction for application programmers to transparently allocate, move, and access data efficiently on heterogeneous memory systems.
Adrian outlined some of the work done in the NEXTGenIO project to exploit persistent memory for computational simulation, data analytics, and machine learning.
Panel session: The future of heterogeneous memories
A panel session on The future of heterogeneous memories closed the event. It was moderated by Marie-Christine Sawley (Intel) and was participated by all the speakers and the audience.
The main discussion during the panel session focused on the need for the user community to realise the power of heterogeneous memory technologies, and how to increase the number of people using them.