The documents distributed by this server have been provided by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a noncommercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.
Publications of SPCL
|Nonblocking and Sparse Collective Operations on Petascale Computers|
(Presentation - . Vol , Nr. , In , presented in Argonne National Laboratory, pages , , ISSN: , ISBN: , Jun. 2010, )
AbstractThis talk introduces new classes of collective operations from an implementation as well as an application programmer's perspective. We discuss issues with schedule generation, caching, and progression, and how these influence the application programmer. Then we focus on simple strategies, such as loop tiling, pipelining, and simple code movement, that can be used to optimize application performance with nonblocking collectives. We also discuss how the new semantics can be utilized to design new, asymptotically optimal algorithms for one-level termination detection, which is important for data-driven algorithms. The second part of the talk focusses on sparse collective operations and static binding of communication topologies. We discuss a possible interface for MPI-3, several productivity and performance issues, and show some performance results and potential for future work and architectures.