This year I will give a talk at the Parallcon in Karlsruhe, Germany. The conference will take place from the 22nd to the 24th of April. My talk will be about an efficient parallelization for the Intel Xeon Phi. Ultimately this will be highly adoptable for other massively parallel systems. Nevertheless, I consider the Xeon Phi the most interesting use-case at the moment. Where else can you handle 60 cores with 240 simultaneously running threads?
Scaling will be one part of the talk. The other one will be efficient prefetching and vectorization. The latter is again more generic, while prefetching is more a less special in the Xeon Phi scenario. In the end this will also introduce the QPACE 2 project, which is in its final stages.