Mini-Workshop: In-Place Updatable Data Structures for UPMEM

We are organizing a mini-workshop on UPMEM to explore its potential 
and future in the context of near-data processing (NDP). It is our 
firm belief that NDP will only gain traction as a standalone technology
if the DRAM it uses can simultaneously store data. Otherwise, using 
NDP devices purely as accelerators requires buying both standard 
DRAM and NDP-RAM, leading to redundant costs. At present, the UPMEM SDK
enforces a transfer model, making UPMEM an accelerator. In this workshop,
we want to discuss data structures and operational modes for UPMEM 
that better balance the roles of host (CPU) and UPMEM (DPU). Thereby,
UPMEM should become usable not only for processing but also for
data storage. We also want to discuss UPMEM's prospects in the upcoming
of CXL-based NDP technologies.

The idea is to do a one-and-a-half day hackathon-like workshop 
with a lots of hands-on-work. We want to present our work 
on a technical level to each other and develop new ideas.
More Details and registration can be found in the pad:

https://pad.gwdg.de/ygbepVqtSm6OlgvX2kyplA?edit