GPU Computing with TESLA™
We conceive GPU Computing as “the” way to deliver parallel computing for the masses as dubbed by Dr Dobbs magazine series of articles. Introducing GFlops-level of performance in everyday use opens a new industrial revolution by changing the scale of complex simulation use. HPC Project axes its business orientation in making such a tremendous technology commonplace for engineers by automating the adaptation of the most used applications for GPU Computing.
GPU Computing delivers the right level of performance for pre & post processing model manipulation as new numerical algorithms contribute to lower the required computing power to deal with complex simulations.
HPC Project offers training services to its main customers to present the concepts within GPU Computing. This allows HPC-Project to help and develop the ecosystem and to demonstrate the competitive advantages of this approach (see the training page). For limited experimentations, HPC Project provides expertise in order to demonstrate the value proposition of its systems. A consulting team is able to service customers for larger projects in training, experimentation and migration.
HPC Project is partnering with high-performance libraries developers to leverage further the computing power delivered by the NVIDIA Tesla board, such as geometry description frameworks, numerical solvers and mathematical & graphics libraries
Use of CUDA
We internally use a parallelizing and compiling framework to generate CUDA code from standard Fortran and C source code. This generic technology is currently expanded to allow a completely automated CUDA code generation. We use CUDA to produce NVIDIA® Tesla™ -accelerated applications.
Our target is every application which makes a strong use of simulation and data-mining. We particularly target widely used applications to integrate them in high-performance appliances equipped with Tesla™ boards (Wild Node™ by Wild Systems® ). This packaging is targeted to small engineering and modelling teams.
We eventually offer larger configurations of multiple nodes connected by a high-performance PCI-express interconnect. In this case, our compiling framework for CUDA code production is made available for the customer in order to facilitate the migration of internal applications towards GPU Computing.
HPC Project is a NVIDIA partner : 
| Top |
