Out-of-core macromolecular simulations on multithreaded architectures

摘要

We address the solution of large-scale eigenvalue problems that appear in the motion simulation of complexrnmacromolecules on multithreaded platforms, consisting of multicore processors and possibly a graphicsrnprocessor (graphics processing unit). In particular, we compare specialized implementations of several highperformancerneigensolvers that, by relying on disk storage and out-of-core techniques, can in principle tacklernthe large memory requirements of these biological problems, which in general do not fit into the mainrnmemory of current desktop machines. All these out-of-core eigensolvers, except for one, are composed ofrncompute-bound (i.e., arithmetically intensive) operations, which we accelerate by exploiting the performancernof current multicore processors and, in some cases, by additionally off-loading certain parts of the computationrnto a graphics processing unit accelerator. One of the eigensolvers is a memory-bound algorithm, whichrnstrongly constrains its performance when the data is on disk. However, this method exhibits a much lowerrnarithmetic cost compared with its compute-bound alternatives for this particular application. Experimentalrnresults on a desktop platform, representative of current server technology, illustrate the potential of thesernmethods to address the simulation of biological activity.