Xavier Andrade edited Parallelization.tex  over 9 years ago

Commit id: 99c084116981384e77381dce402af2792feceaae

deletions | additions      

       

In the development of new theory, computational cost is also an important factor.  Real-space grids allows us to provide developers of new method withoptimized and parallelized  building blocks that they can use to produce highly efficient code without caring about the details of the implementation. In most cases this blocks allow developers to write code that is automatically parallel, efficient and that can run in massively parallel processors like graphics processing units (GPUs).  However, it is critical to expose an adequate level that hides the performance details while still giving enough flexibility to the developers. For example, we have found that the traditional picture of an state as the basic object is not adequate for optimal performance, as it does not expose enough data parallelism~\cite{Andrade_2012_gpus}. In Octopus we use a higher level interface where the basic object is a group of several states.