A propos du centre ou de la direction fonctionnelle

The Centre Inria de l'Université de Grenoble groups together almost 600 people in 22 research teams and 7 research support departments.

Staff is present on three campuses in Grenoble, in close collaboration with other research and higher education institutions (Université Grenoble Alpes, CNRS, CEA, INRAE, ...), but also with key economic players in the area.

The Centre Inria de l'Université Grenoble Alpe is active in the fields of high-performance computing, verification and embedded systems, modeling of the environment at multiple levels, and data science and artificial intelligence. The center is a top-level scientific institute with an extensive network of international collaborations in Europe and the rest of the world.

Contexte et atouts du poste

CORSE is a joint research group in the LIG laboratory that regroups several expertise that stand at the interface between software and hardware: those are domain specific application/library tuning, compiler optimization, run-time systems, and debugging/monitoring. Our domains of application include performance (both speed and energy consumption), reliability, and teaching of computer science. An important activity concerns the optimization of machine learning applications for some specific high-performance embedded architectures.

The position is funded by the Holy Grail project. The goal of CORSE in this project is to advance research in compiler optimization, including compiler infrastructure and scheduling heuristics, specifically for reduced deep learning codes such as those that are quantized or pruned.

Mission confiée

The aim of the position is to contribute to the development of tools that help the programmer to obtain highly optimized code of deep learning applications that have been pruned. We focus on a limited number of operators: matrix multiplication, matrix-vector multiplication, convolution. This work includes various tasks:

• Background: understanding of existing methods in state-of-the-art compilers (typically MLIR)
• Language design : conception and implementation of the primitives which describe the sparse computations, including decompression ;
• Efficient code generation, using traditional techniques (tiling...) but also pattern-based vectorization of such operators, exploiting the fact that the sparse matrix is statically known. See for example : https://www.cs.toronto.edu/~mmehride/papers/PSC.pdf

The targeted architectures are CPUs.

Principales activités

The main activities include:

• Experiment with different optimization schemes on various CPU architectures
• Conception of various programming primitives
• Develop a code generator

Compétences

The position requires:

1. Development experience with MLIR
2. Background in compiler infrastructures for deep-learning applications
3. Expertise in compiler optimization focusing on data locality and parallelism (including data dependencies, tiling, etc.)
4. Proficiency in C, C++, and Python programming
5. Strong communication skills (teamwork) and the ability to thrive in a research environment with flexible development directives

Avantages

• 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
• Social, cultural and sports events and activities
• Access to vocational training
• Social security coverage under conditions

Rémunération

Gross salary : from 2 692 euros before deduction of tax incomes depending on laboral experiences and degrees.