文档介绍:CC-MPI: munication Capable MPI Prototype for Switched ClustersAmit Karwande, Xin YuanDepartment puter Science, Florida State UniversityDavid K. LowenthalDepartment puter Science, University of ia?Motivation?Related -MPI–One-to-many(all) communications –Many(all)-to-many(all) communications?Performance study?Conclusion?munication libraries (. MPI) work details and provide a simple API.?Advantage: User friendly.?Limitation: Communication optimization opportunity is limited.–Optimizations can either be done at piler or in the library.?Architectural independent optimizations in piler?Architectural dependent optimizations in the library, but such optimizations can only be done for a single routine.?munication:–pile time, use both the munication work architecture information to munication optimizations.?Static management work resources.?Compiler directed architecture dependent optimizations.?Architecture dependent optimization across patterns.–To apply munication technique to MPI programs, ?The library must closely match the MPI library.?The library must be able to support optimizations munication.–work details.–Different implementations for a routine so that the user can choose the best one.–This work focuses on munication munication library.?Related Work:–Compiled directed architectural dependent optimization [Hinrichs94]–munication [Bromley91,Cappello95,Kumar92,Yuan03]–MPI optimizations [Ogawa96,Lauria97,Tang00,Kielmann99]?CC-MPI:–Optimizes one-al-all, one-to-many, all-to-all, and many-to-munications–Targets Switched Clusters–Basic idea:?work control routines from data transmission routines?Multiple implementations for each MPI routine.?One-to-many(all) communications:–Multicast based implementations.?Reliable multicast (IP multicast is unreliable)–Use a simple ACK-based protocol)?Group Management–A group needs to be created before munication is to be performed.–2^n potential groups for n members–The hardware limits the number of simultaneous groups.–CC-MPI suppo