| < Previous | Other Sessions | Next > |
Organizers:
Luciano Garelli
Nicolas Wolovick
Description:
The continuous growth in the computational capabilities of computers requires the adaptation of the way of working in order to optimally take advantage of the computational resources. Since 2005, the only argument that has increased the speed of processing in computers has been the parallel calculation, either parallelism of instructions, data, threads or nodes. The complexity of current computers that involve these four forms of parallelism, requires that the programming efficiently take advantage of these billions of transistors to obtain good performance, efficiency and scalability. This also leads to proposing optimizations to programs in use, as well as new algorithms and data structures for problems in computational mechanics and multiphysics. This session tries to provide a forum for those working in the resolution of computational multiphysics and mechanics problems in traditional processors (CPU) or accelerators (GPU / MIC), taking advantage of all its dimensions of parallelism, as well as in clusters of nodes of these processors.
KEYWORDS: