Coupling a pattern-matching language to a source-to-source compiler¶
Internship for a MSc or engineer student
- Keywords: pattern-matching language, compilation, parallelization,
- program transformation, optimization
Current computing architectures are all evolving toward parallel architectures with always more processors for power-efficiency reasons: processors with 8+ cores, GPU for graphics and computations with 1600 cores (for example the old 2.7-TFLOPS AMD/ATI HD 5870)… This means that the stress is put on the programmer to parallelize her application to such an extend to fit the target application. Unfortunately, this is quite more difficult than plain sequential programming.
To help the programmer job, many tools are developed: new parallel languages, new parallel libraries, new autoparallelizer compilers for existing languages, etc.
In this internship, we focus on autoparallelizer compilers with the ability to find out in the programmer source code some constructions that can efficiently be executed on parallel architectures with specific instructions (SSE intrinsics for example) or libraries (optimized linear algebra function such as a matrix multiplication).
The pattern matching is an old research domain we don’t want to reinvent and the aim of the internship is to find the best pattern-matching tool (http://tom.loria.fr seems a good candidate…) and language suitable to our domain and to couple it with ROSE Compiler to have a powerful source-to-source compiler with advanced algorithm recognition system.
Advisor: Ronan Keryell rk(at)hpc-project(dot)com
Location: HPC Project http://www.hpc-project.com which is a fast-growing start-up with 35 people at Meudon (92) and Montpellier (34), France and Santa Clara (USA).
