Parallel Numerical Linear Algebra Researchers Meet in Manchester

Some of the NLAFET team
Some of the NLAFET team

Sixteen researchers from the project NLAFET: Parallel Numerical Linear Algebra for Future Extreme-Scale Systems met in the School of Mathematics for two days earlier this month to review progress and plan future work. This €2.8M EU-funded project is led by Umeå University (Sweden), with partners Manchester, INRIA (Paris), and the Rutherford-Appleton Laboratory. It aims to develop linear algebra software for future extreme-scale systems that will enable radical improvements in the performance and scalability of a wide range of real-world applications.

The Manchester component of the project is led by Professors Jack Dongarra and Nick Higham and employs three postdoctoral researchers. Together with colleagues in Knoxville the Manchester team is developing a new standard for parallel evaluation of multiple small low-level linear algebra operations, motivated by applications in astrophysics, fluid dynamics, and machine learning. Comments from the community on the initial documents;

A Proposed API for Batched Basic Linear Algebra Subprograms 

Workshop on Batched, Reproducible, and Reduced Precision BLAS 

A Comparison of Potential Interfaces for Batched BLAS Computations

are welcome.

▲ Up to the top