PhD Position F/M Dynamic Parallelization of Sparse Codes for Machine Learning and High-Performance Computing

at Inria
Published March 11, 2023
Location Lyon, United States of America
Category Machine Learning  
Job Type Scholarship  

Description

Contract type : Fixed-term contract

Level of qualifications required : Graduate degree or equivalent

Fonction : PhD Position

About the research centre or Inria department

The Inria research centre in Lyon (previously the Lyon branch of the Inria centre in Grenoble) is the 9th Inria research centre, formally created in December 2021.  It brings together approximately 270 people (including 110 Inria employees) in 15 research teams and research support services.

Its staff are distributed at this stage on 2 campuses: in Villeurbanne La Doua (Centre / INSA Lyon / UCBL) on the one hand, and Lyon Gerland  (ENS de Lyon) on the other. A third site should be opened in the course of 2022. The teams are mainly hosted with our partners.  The centre's teams work closely with research and higher education institutions (ENS de Lyon, UCBL, INSA Lyon, etc.), their laboratories, and other research organisations in Lyon (CNRS, INRAE, competitiveness clusters, etc.), but also with Lyon and regional economic players. Many international collaborations are also underway.

The Lyon centre is active in the fields of software, distributed and high-performance computing, embedded systems, quantum computing and privacy in the digital world, but also in digital health and computational biology.

Context

This PhD thesis will be held at Inria Lyon, located at LIP laboratory (ENS de Lyon), France in collaboration with :

  • Philippe Clauss (Inria Strasbourg, CAMUS team)
  • Thierry Gautier, (Inria Lyon, AVALON team)
  • Xavier Rival, (Inria Paris, ANTIQUE team)

LIP is a renewed French research laboratory covering a wide spectrum of key topics of computer and information sciences as well as on various inter-disciplinary initiatives. LIP main strength is the creative interaction between long-term fundamental research, innovative software/hardware design and shorter-term projects/transfers through industrial collaborations.

Assignment

Since the early days of parallel computing, industry is pushing towards programming models, languages and compilers to help the programmer in the tedious task to parallelize a program.  Automatic parallelisation focuses on programming automatically parallel computers from a sequential specification. In the past decades, a general unified framework, the polyhedral model, was designed to solve that problem for regular loop kernels manipulating dense tensors (arrays).  With the polyhedral model, compilers may reason about programs at iteration-level, giving rise to powerful automatic parallelization algorithms. However, most kernels of interest (high-performance computing, machine learning) manipulate sparse tensors.

The overall goal of this PhD thesis is to extend the scope of automatic parallelization to sparse tensors.

Main activities

Specifically, this PhD thesis will investigate how to delay the optimization of sparse code at runtime when the sparse structure is known. Given the dense specification, we aim at producing a code able to specialize itself on the input sparse structure, resulting in a parallel code using state-of-the art linear algebra libraries}. Many issues and trade-offs must be investigated. To quote a few:

  • How to represent / retrieve properly the sparsity ? How propagate the sparsity along the computation flow?
  • How to specialize the code with library kernels?
  • How to enforce a proper scheduling for the parallel runtime?

The PhD student will address all these questions and validate his approach on scientific benchmarks by using
sparse tensors from the Florida sparse matrix collection; as well as machine learning applications.

 

Skills

Notions in compilers, parallelism, parallel architectures. Experience with C++.

Benefits package

  • Subsidized meals
  • Partial reimbursement of public transport costs
  • Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
  • Possibility of teleworking (90 days / year) and flexible organization of working hours
  • Professional equipment available (videoconferencing, loan of computer equipment, etc.)
  • Social, cultural and sports events and activities
  • Access to vocational training
  • Social security coverage under conditions

Remuneration

1st and 2nd year: 2 051 euros gross salary /month

3rd year: 2 158 euros gross salary / month