Main interests

- Accelerators & Hybrid Processors
- Manycore Architectures
- Task Scheduling & Data Management
- Runtime Systems
- Parallel Programming
- High Performance Computing


I am currently a postdoc at the French department of Energy (CEA) where I work on large-scale clusters of machines equipped with accelerators. Before, I was in the CUDA team at NVIDIA. Prior to this, I was a PhD student at the University of Bordeaux. You can download my CV in this page.

Read More


During my PhD thesis, I've designed the StarPU runtime system which schedules tasks and automates data management within heterogeneous multicore platforms enhanced with accelerators.

Read More

Hits From The Blog

Computing pi by rolling (many many) dices

Here is a rather unexpected way to compute pi that I've sometimes encountered when looking for massively parallel algorithms. As a foreword, if you are trying to compute the billion-th digit of pi, you can stop reading this right now This technique is part of a  class of algorithms called "Monte Carlo method" which consists [...]

CUDA kernels launches in the null stream are NOT synchronous

Today I'd like to point out some very common error found in CUDA codes which assume that calling a kernel with stream 0 (either explicitly or implicitly) will result in a synchronous kernel call. Prior to the introduction of streams in CUDA, users did not have to care about synchronization issues, as everything would typically [...]


Subscribe to our e-mail newsletter to receive updates.