Porting Parallel STL- style libraries to heterogeneous processors

Affiliated to

Location

Timing

Description

The STL (Standard Template Library) is a collection of commonly used algorithms and data structures for the C++ Programming language. This project involves adapting existing parallel STL- style libraries for heterogeneous processors such as the Cell Processor and/or GPUs using Codeplay's offload technology. This is an opportunity to work with the latest compiler technology to research and create new parallel software libraries that will be made available to developers for free.

Libraries likely to be considered for the projects are:
- libstdc++ parallel mode http://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
- HPC++ http://www.extreme.indiana.edu/hpc++/index.html
- STXXL (STL for large datasets) http://stxxl.sourceforge.net/

Duties will include offloading parallelised code in these libraries using Codeplay's Offload compiler (http://offload.codeplay.com/); writing, profiling and tuning test applications; identifying further opportunities to improve performance through offloading. A researcher working on this project will learn about how to achieve high levels of performance and memory bandwidth on real C++ applications.

Codeplay is a small company (11 staff), working in a high-tech area (multicore optimization for games and media processing) there will be plenty of opportunities to learn about new processor and software technologies. You will also be working directly with staff working on games and compiler projects for advanced processor architectures.

Requirement for this internship is a solid understanding of:
- C++ and template libraries
- multithreading
- software optimization

(There are 2 positions available on this project. The project can be divided easily into working on different libraries or components)