Type of Position: Research Fellowship (Bolsa de Investigação)
Type of Contract: Research grant
Duration: 7 Months
Closed at: 2021-Feb-25
This grant is offered in the context of the European project EPEEC, "European joint Effort toward a Highly Productive Programming Environment for Heterogeneous". The main general objective of EPEEC is to develop and deploy a well-integrated, comprehensive parallel programming environment targeting exascale systems that turns next-generation supercomputers into manageable platforms for domain application developers. One important component of this project is the development of weakly consistent collectives that achieve better performance by relaxing some of the semantics of these primitives. The goal of this work is to develop a collective for the GASPI platform, based on allreduce, geared towards machine learning computations, and that allows us to explore the design space of state of the art machine learning techniques, including but not limited to the use of a parameter server, and bounded staleness processing. This collective will then be applied, in a first step, to a drug discovery application that uses BPMF, and, time permitting, to deep learning.
Rodrigo Seromenho Miragaia Rodrigues