Engineering Services

During the 80's, supercomputing was a niche market, dominated by companies building hyper-specialized architectures, for the benefit of scientific research centers. Today, a new generation of processor architectures has emerged, and leads to a technological and economical leap, depending on the capacity of software codes to fully use the power of such processors.

In this context, the ability to adopt a strategy in terms of parallel software architecture will build an "entry barrier" for competition, and ensures software products will remain competitive with respect to operational requirements. However, this strategy often requires to re-engineer existing codes, to fully use the capacities of this last generation of processors. To address this issue, HPC Project develops strategies allowing, without extensive investment in a re-engineering project, the full use of legacy applications and already developed software.

Services

• HPC Project can design both software and hardware architectures for simulation and High Performance Computing. HPC Project team has an extensive experience in developing and designing electronic circuits, and is able to assist its customers from the specifications until the satisfaction of initial requirements.
• HPC Project performs code analysis and optimization. Experts from HPC Project are able to use both platform independent and platform dependent techniques. While the first are generic techniques that are effective for most platforms, our strength relies in the mastering of very specific cutting-edge platform dependent techniques which involve instruction level parallelism, data level parallelism, cache optimization techniques, etc...

To enable such capabilities, our framework has built-in features to the source code transformation:

• Parallelizing compiling schemes and inter-procedural analysis featuring detection of candidate hotspots, reorganization of loops, preparation of data alignment, cache fetching, load balancing, software pipelining organization and kernel codes stubbing
• Formal intermediary representation
• Back-end targeting (CUDA for TESLA Boards)
• Speculative code production

Contact us to discuss the specifics of your project.

Return to the "Offer" page