Post-Doctoral Research Visit F - M Performance Analysis Of ai Workloads H/F

INRIA

Post-Doctoral Research Visit F/M Performance analysis of AI workloads
Le descriptif de l'offre ci-dessous est en Anglais
Type de contrat : CDD

Contrat renouvelable : Oui

Niveau de diplôme exigé : Thèse ou équivalent

Fonction : Post-Doctorant

Niveau d'expérience souhaité : Jeune diplômé

A propos du centre ou de la direction fonctionnelle

The Centre Inria de l'Université de Grenoble groups together almost 600 people in 26 research teams and 9 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 Alpes 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
As part of an innovative research project, we are seeking a post-doctoral researcher for a one- year position, renewable. The context is compiler optimization for AI compute kernels. The focus is on performance analysis and modeling, both for optimization purposes and performance debugging. In this context, the team is developing several tools. The first tool concerns the automatic characterization of microarchitectures, covering all stages that impact performance such as instruction decoding, branch prediction, scheduling queues, and cache management policies at various levels. The second tool focuses on performance prediction and debugging of compute kernels. The goal here is either to quickly predict (without compiling and executing) which version of a kernel is faster, or to identify what is slowing down (e.g., instructions, dependencies, cache misses) the execution of a compute kernel. The use of these tools is specifically focused on the super-optimization of deep learning programs within the Holigrail (PEPR IA), Deepgreen (BPI), and Camelia (PEPR agency) projects. Within these projects, CORSE aims to contribute to the development of software and hardware infrastructures to improve the efficiency of deep neural networks. The objective is particularly to develop a modern, toolbox-style compilation infrastructure, providing expert programmers with the tools needed to automate the super-optimization of their compute kernels.
Mission confiée
Several PhD students and engineers contribute to the tools mentioned above. The objectives are multifaceted. The first involves developing new techniques for the automatic characterization of microarchitectures to improve prediction accuracy. The second focuses on extending the GUS tool (a cycle-approximate microarchitecture simulator and performance analyzer/debugger), particularly by making GUS's raw data more readable, understandable, and explorable for developers. This will be achieved by enhancing its user interface with contextualized, interactive, and text-based presentations. Additional extensions are planned, such as enriching GUS outputs by merging them with outputs from tools like Perf. The postdoc's mission will be to contribute to the theoretical development of these new techniques and to support and supervise the students for whom this is their primary project.
Principales activités

Co-supervision of PhD students on the topics described above

- Participation in technical discussions
- Literature review and bibliographic work
- Co-development of tool extensions such as PALMED, GUS, or perf
- Co-authoring of scientific papers
- Attendance and participation in seminars/conferences related to the position

Compétences

- C/C++ proficiency
- Compiler technologies background
- Performance analysis and binary translation, espetially using tools such as perf, GEM5, ...
- Spoken and written english

Avantages

- Subsidizedmeals
- 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

2788 euros gross salary /month