Abstract
We discuss the implementation of an algorithm for the generalized eigenvalue problem on a distributed-memory, hybrid multiprocessor. The basic goal is to obtain good concurrency, data reference locality, load balance and a task granularity that is consistent with the expected parallelization overhead. The implementation is based on restructured EISPACK routines using various standard high-level parallel constructs such as serieal sections, parallel do-loops, critical sections and barriers. In addition, we also consider the explicit use of dynamic task scheduling at the application level to perform the highly concurrent management of work queues. The issues related to algorithm development and performance are discussed and speedup and efficiency measurements are presented.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.