Speaker
Mr
Peter Georg
(University of Regensburg)
Description
On many parallel machines, the time LQCD applications spend in communication is a significant contribution to the total wall-clock time, especially in the strong-scaling limit.
We present a novel high-performance communication library that can be used as a de facto drop-in replacement in existing software.
Its lightweight nature that avoids some of the unnecessary overhead introduced by MPI allows us to improve the communication performance of an application without any algorithmic or complicated implementation changes.
As a first real-world benchmark, we make use of the library in the coarse grid solve of the DD-\alphaAMG algorithm.
On realistic lattices, we see an improvement of a factor 2x in pure communication time and total time savings of up to 20%.
Author
Mr
Peter Georg
(University of Regensburg)
Co-authors
Mr
Daniel Richtmann
(University of Regensburg)
Prof.
Tilo Wettig
(University of Regensburg)
