ELPA - Algorithmic extensions and optimization

The solution of symmetric eigenproblems plays a key role in many computational simulations. Especially in ab-initio molecular dynamics it is by far the most expensive part. ELPA is one of the leading libraries in the computation of parts or the whole eigenspektrum on large scale compute clusters. The project finished in 2011 and some progress was still going on till now. Based on comparisons with other parallel solvers we will discuss future improvements of the ELPA code. A further goal is to exploit the computing power of modern heterogeneous architectures. Additionally we will present further algorithmic ideas that might improve the performance and the applicability of the ELPA library.     «

The solution of symmetric eigenproblems plays a key role in many computational simulations. Especially in ab-initio molecular dynamics it is by far the most expensive part. ELPA is one of the leading libraries in the computation of parts or the whole eigenspektrum on large scale compute clusters. The project finished in 2011 and some progress was still going on till now. Based on comparisons with other parallel solvers we will discuss future improvements of the ELPA code. A further goal is to e...     »