Numerical Linear Algebra and Benchmarking for Supercomputing Hardware


Wednesday, 10 January, 2018 - 17:00


This talk will present an international landscape of supercomputing
hardware  from the perspective of decade-old development of production
quality numerical libraries. In particular, large supercomputing
installations and scientific clusters will be discussed with the
emphasis on their architectural features that are essential for good
performance and scalability. This will give an appropriate context for
the design, implementation, and optimization of High Performance
Computing libraries for high quality scientific software infrastructure.
This will lead to the performance assessment and engineering aspects
and, in particular, some modern benchmarks of prominence, including HPL,
HPCC, and HPCG. Their influence on these libraries will also be

Lecturer: Piotr Luszczek from the University of Tennessee