2-year Post-doctoral Position: Open Hardware - Rennes, France - INSA RENNES

INSA RENNES

Entreprise vérifiée

Rennes, France

il y a 3 semaines

Posté par:

Sophie Dupont

beBee Recruiter

Description

2-year Post-doctoral Position - Open hardware compilation for near-memory dataflow computing

Keywords:
High-Performance computing, Near-memory computing, Compilation, Dataflow models of computation ,
Stream processing, Open hardware, RISC-V, FPGA, Computer architecture.

Laboratory:
Institut d'Electronique et des Technologies du numéRique

UMR CNRS 6164, Rennes, France

Context

Near memory computing (NMC) consists of adding computing capacities (typically processors) as close as possible to the main memory of the computer, thus bypassing the hierarchy of caches, and limiting its numerous data transfers
[Singh19]. Memory is a vital part of any digital system. It is used to store the programs of the system and the data on which these programs operate. The speed at which a program runs directly depends on the speed at which instructions and data can be transferred between memory and processor. The memory must therefore be (very) fast, but ideally also (very) large to store all programs and data, and inexpensive. The impossible meeting between these three requirements has led to proposing solutions based on a layered memory hierarchy, relying on various memory technologies. Through the numerous hardware solutions implemented from generation to generation, more and more transistors and energy in a circuit are allocated for the sole purpose of improving memory access [Horowitz14]. Cache memories, embedded as close as possible to the processor on the same chip gives the programmer the illusion of both fast and large memory. The cache memory exploits the assumptions of both spatial and temporal data access locality, assumptions that do not hold when processing data streams [Ghasemi21], as data undergo a
- process-and-forget_ management. This post-doctoral contract is financed by the
RIDIM project, led by Lab-STICC and involving Irisa and IETR laboratories, in collaboration with the University of Maryland, USA. The RIDIM project aims to integrate stream processing capabilities all along the data path from the main memory to the processor.

Objectives
PiSDF dataflow functional descriptions [Desnos13].

Environment and Expected Impact
The post-doc will benefit from a close collaboration between the three RIDIM research teams: the project is coordinated by Kevin
Martin at Lab-STICC in Lorient and involves Steven Derrien from the TARAN team of IRISA and the VAADER team of IETR in
Rennes.

The project will build on the momentum of the Open Hardware and RISC-V communities that are eager for novel architecture propositions.

RIDIM teams are involved in this community and participate to the organising committee of the yearly RISC-V week, largest event on RISC-V in Europe .

The RIDIM post-doc will be closely linked to the Cominlabs
International Chaire at INSA/IETR of Prof. Shuvra S. Bhattacharyya, Professor of Electrical and Computer Engineering at the

University of Maryland, College Park, USA. The post-doc will be located in Rennes, France, a very dynamic city with nearly
700,000 inhabitants under its urban umbrella, and a fast growing population (the city has the second largest population growth in
France).

This post-doc and the RIDIM project are expected to set the foundations of a programming model and its supporting hardware for building smarter distributed systems with stream processing facilities.

The main objective of project dissemination is through high quality publications at events such as DATE, DAC, Embedded Systems Week, ACM TACO, as well as in highly ranked journals on system design.

Required experience Experience to be gained in the post-doc
Software design (required), Compilation (appreciated) Open source hardware and RISC-V systems
C, C++, and (Java or LLVM/MLIR) (required) Dataflow models of computation
Python (appreciated) Project-level task lead

Astronomy computing workloads
Digital hardware design (appreciated)
Parallel programming (appreciated)

Characteristics location:
Vaader team, IETR laboratory - INSA Rennes, 20 Avenue des Buttes de Coësmes, 35708 Rennes, France

Duration: 2 years, Start: ASAP

Salary per month:
depends on experience, minimum 2950€ gross / 2350€ netto
Supervisors

Maxime Pelcat (IETR, INSA Rennes) - [at]insa-Jean-François Nezan (IETR, INSA Rennes) - jnezan[at]insa-Karol Desnos (IETR, INSA Rennes) - kdesnos[at]insa-
Applications

References
[Desnos13] Karol Desnos, et al

"Pimm:

Parameterized and interfaced dataflow meta-model for mpsocs runtime reconfiguration."
SAMOS XIII. IEEE, 2013.
[Horowitz14] Mark Horowitz. "Computing's energy problem (and what we can do about it)"". In 2014 IEEE ISSCC. IEEE, 2014.
[Kermarrec19] Florent Kermarrec, et al

"LiteX:

an open-source SoC builder and library based on Migen Python DSL." OSDA
2019.
[Lattner21] Chris Lattner, et al

"Mlir:

Scaling compiler infrastructure for domain specific computation." 2021 IEEE/ACM CGO.
IEEE, 2021.
[Miomandre20