In parallel to our libraries, we offer our services to carry your applications to the GPU in order to make parallelisable HPC applications. The final application will use the GPU in an automatic and transparent manner and you can then experience considerable time gains.
The carrying of your applications to GPU is realised directly in the graphic card’s assembler language in order to optimise performance. The application is globally parallelised in order to prevent bottlenecks, which could reduce performance.
The time needed for development depends on the application concerned. Our GPU-Computing service offer is carried out in 3 phases:
Understanding of the concerned application and research on the parts of the algorithm using floating calculations and generating the most computing times, analysis of parallelisation possibilities and organisation of data.
