论文信息 - Design, Implementation, and Usage of LibNBC

Design, Implementation, and Usage of LibNBC

We describe the design and implementation of LibNBC a library that implements nonblocking collective operations. Its main goals are high portability and high performance. The library is written in ANSI C on top of MPI-1. This document describes the internal design, implementation and various internal and external programming interfaces of LibNBC.

Torsten Hoefler | Andrew Lumsdaine

[1] Torsten Hoefler,et al. Fast barrier synchronization for InfiniBand , 2006 .

[2] Sergei Gorlatch,et al. Send-receive considered harmful: Myths and realities of message passing , 2004, TOPL.

[3] George Bosilca,et al. Open MPI: Goals, Concept, and Design of a Next Generation MPI Implementation , 2004, PVM/MPI.

[4] Keith D. Underwood,et al. An analysis of the impact of MPI overlap and independent progress , 2004, ICS '04.

[5] Dipl.-Inf. Torsten Hoefler,et al. A Survey of Barrier Algorithms for Coarse Grained Supercomputers , 2005 .

[6] Debra Hensgen,et al. Two algorithms for barrier synchronization , 1988, International Journal of Parallel Programming.

[7] J. White,et al. An Analysis of Popular Mpi Implementations , .