Agenda

Abstract: Computer clusters provide high computational power and high scalability but for medium/high size systems the maintenance and power consumption costs become intolerable. Heterogeneous architecture systems, where a CPU coexists with an accelerator that enable massive parallelization, arise as an attractive solution to this problem.  They provide high-perfomance computational capabilities as well as reduced power consumption, yet the design methodologies are more complex and time costly than the traditional software development. FPGA devices stand out from the different available alternatives. This technology enables a complete customization of the computing architecture leading to extremely high speedups in comparion to traditional CPUs. For some cases, it is even possible to replace a whole PC cluster with a single FPGA device, thus, reducing power consumption drastically. In this talk, two design examples that lead to low-cost high-performance computing are explained: DNA sequence alingment and computational fluid dynamics. Both examples are extreme cases that require very different design methodologies and they also achieve different speedups: x1000 for DNA alignment and x60 for CFD.

L'équipe CAIRN vous invite à ce séminaire public qui se tiendra dans les locaux de l'Enssat à Lannion (22), en salle 036C.