I float like a butterfly and sting like a bee.
— Muhummad Ali
I am currently working as full-time R&D engineer at Quarkslab, having fun hacking C, C++ and Python compilers.
I am interested in many research fields and try hard to focus on:
- automatic code obfuscation
- (source-to-source) compiler infrastructures
- compilation of scripting languages (esp. Python)
- efficient simd code generation (sse, neon, avx and the likes)
- parallel runtimes / libraries
My PhD's research subject was building source-to-source compilers for heterogeneous machines. I am very grateful to my advisors, François Irigoin and Ronan Keryell with whom I shared a really nice piece of scientific (and more) life.
Most of my PhD activities focused on the enhancement of the PIPS project. I contributed to
- PyPs, a programmable pass manager for PIPS
- transformations for accelerator code generation: isolate statement, delay communications, symbolic tiling, memory footprint estimation etc
- retargetable multimedia instruction compiler
- an iterative compiler, a.k.a. PIPS fuzzer
- and some other transformations: outlining, inlining, instruction selection etc
I also tried hard to improve (?) some engineering aspects of PIPS, mainly by porting the build infrastructure to autotools and enforce the use of PyPs.
I have been involved in the following projects, and as time flies I am still in touch with them :p
- quiet-oceans, a firm specialized on oceanic noise forecasting. I did a lot of HPC there!
- ramsurf, a noise propagation model I ported and accelerated from FORTRAN to C
- hyantes, a geomatic tool to compute the influence of potentials in one point on the neighborhood
- kaapi, a work-stealing base middle-ware for distributed computing.