Access more resources about UPMEM

We have gathered a selection of past and on-going works related to UPMEM’s technology, often involving renowned Universities with links to access to papers, videos and open source code when available .

For clarity purposes sections are sorted by fields of interest.

Last updated on April 2022

About the architecture, general performances and micro-benchmarks

Benchmarking a New Paradigm: An Experimental Analysis
of a Real Processing-in-Memory Architecture
Published ( 2021)ETH ZurichFull paper
GitHub repo
Benchmarking Memory-Centric Computing Systems: Analysis of Real Processing-In-Memory HardwarePublished (CUT 2021)ETH ZurichPaper
GitHub repo
A Case Study of Processing-in-Memory in off-the-Shelf Systems
Published (USENIX 2021)University of British ColumbiaPaper
Full GitHub
Architecture programming tools overview, showcasing genomics/analytics applications using UPMEM Processing In MemoryPublished (HIPEAC 2021)UPMEMVideo
SDK Documentation
DPU demo codes
A survey on hardware accelerators: Taxonomy, trends, challenges, and
Published (JSA 2022)University of Sienna, HuaweiPaperN/A

Genomics & Bioinformatics

Variant Calling Parallelization on Processor-in-Memory Architecture
Published (BIBM 2021)IRISA / CNRS, INRIAPaperGitHub
BLAST software on the UPMEM architecturePublished (2016)INRIAPaperUpon request
High-throughput Pairwise Alignment with the
Wavefront Algorithm using Processing-in-Memory
Published (Hi Comb 2022)ETH ZurichAbstractAvailable soon

Analytics & Database

Index SearchCompletedUPMEM / Private companyUpon requestUPIS repo
PIM-Embedding-LookupIn progressUniversity of British ColumbiaOn-goingGithub
PIM Swap: A page swapper using PIMIn progressUniversity of British ColumbiaOn-goingGithub
Database acceleration on PIM architectureIn progressCarnegie Mellon University On-goingAvailable soon
SparseP: Towards Efficient Sparse Matrix Vector
Multiplication on Real Processing-In-Memory Systems
CompletedETH ZurichPaperGithub repo

Machine Learning / Deep Learning

Evaluating Machine Learning Workloads
on Memory-Centric Computing System
Published (ISVLSI-22)ETH ZurichAbstractKmeans Repo
CNN implementation strategy on a specialized PIM architectureOn-goingENSIMAG / TIMAEarly release upon requestAvailable soon
An implementation of a deep learning recommendation model (DLRM)On-goingUniversity of British ColumbiaAvailable soonGithub