Julien Pagès
Biography
I am a PhD in computer science specialized in compilers, I'm currently working as as software engineer. I am interested in the design and implementation of programming languages, especially with just in time compilation. I did my PhD in the University of Montpellier under the direction of Roland Ducournau. I worked on a virtual machine for a research Object-Oriented language called Nit. The research was focused on implementing efficiently Object-Oriented mechanisms in the context of a JIT compiler and in multiple inheritance. This was a collaboration with the Université du Québec à Montréal (UQAM) in Montréal, Canada. Then I worked for one year in the INRIA team CAMUS in Strasbourg, France on automatic parallel optimizations for dynamic languages. We used polyhedral optimizations and tools which are usually used on static languages like C on the JavaScript language. We implemented a proof of concept in the JavaScript Core virtual machine of the webkit project. From september 2017 to December 2018 I was a postdoctoral fellow in the Université de Montréal and I worked with Marc Feeley on implementing dynamic languages. During this project I have develop a proof of concept of a Python JIT compiler based on the Basic-block Versionning technique. After that I worked for Octasic, a private company based on Montréal as a software engineer. Octasic is developing DSP for embedding systems which are then typically used in radio applications. I was helping them to develop and improve their compiler from C to their custom assembly. This compiler was based on LLVM and clang. Since December 2020 I am working for AMD as a Senior Software Engineer. I am working on the open-source shader compiler based on LLVM for current and future architectures of AMD GPUs.
Publications
Conferences
- Julien Pagès, Marc Feeley. Twopy: A Just-In-Time Compiler For Python Based On Code Specialization Virtual Machines and Language Implementations, VMIL'18, 2018 [ slides ]
-
Manuel Selva, Julien Pagès, Philippe Clauss.
Polyhedral Optimization For JavaScript: The Challenges
8th International Workshop on Polyhedral Compilation Techniques, IMPACT'18, 2018. [ paper | slides ] -
Julien Pagès, Roland Ducournau, Jean Privat.
Preexistence and Concrete Type Analysis in the Context of Multiple Inheritance
International Conference on Principles and Practices of Programming on the Java Platform: virtual machines, languages, and tools, PPPJ'16, 2016. [ paper | slides ] -
Roland Ducournau, Julien Pagès, Jean Privat, Colin Vidal.
Preexistence revisited
Implementation, Compilation, Optimization of Object-Oriented Languages, Programs and Systems Workshop, ICOOOLPS'15, 2015. [ paper | slides ] -
Julien Pagès
A virtual machine for testing compilation/recompilation protocols in multiple inheritance
ECOOP Doctoral Symposium '15, 2015. [ paper | slides ] - Roland Ducournau, Julien Pagès, Jean Privat, Colin Vidal. Genericity and (Co)variance, an Empirical Study. LIRMM, Université de Montpellier, 2014
PhD and Master thesis
-
Julien Pagès sous la direction de Roland Ducournau.
Une machine virtuelle en héritage multiple basée sur le hachage parfait
LIRMM - Université de Montpellier, Thèse de doctorat, 2016 [ PhD | slides ] -
Julien Pagès.
Étude de machines virtuelles Java existantes et adaptation au hachage parfait
Mémoire de master 2 recherche, LIRMM - Université de Montpellier, 2013 [ Master | slides ]
Talks and seminars
University teaching
Teaching assistant (for around 415h) in Universities of Montpellier and Strasbourg
University of Strasbourg : 2016-2017
- Introduction to algorithmic and programming, 56 hours
- Introduction to databases, 38 hours
- Network programming, 34 hours
- Network architecture, 34 hours
University of Montpellier : 2012-2016
- Advanced object-oriented programming in C++, 21 hours
- Modelling and object-oriented programming, UML and Java, 61,5 hours
- Introduction to algorithmic and programming, 51 hours
- Imperative programming in C, 70,5 hours
- Basic concepts in computer science, Linux, 48 hours
Contact
julien.m.pages(at)gmail.com
