Electrical Engineering
      and Computer Sciences

Electrical Engineering and Computer Sciences

COLLEGE OF ENGINEERING

UC Berkeley

Implementing an Efficient Portable Global Memory Layer on Distributed Memory Multiprocessors

Stephen S. Luna

EECS Department
University of California, Berkeley
Technical Report No. UCB/CSD-94-810
May 1994

http://www.eecs.berkeley.edu/Pubs/TechRpts/1994/CSD-94-810.pdf

Libsplit-C is a library of communication and global storage primitives for distributed memory multiprocessors. The library has been implemented on the CMAM and CMAML active message layers for the CM5, the HPAM active message layer for an FDDI network of HP workstations, the NX message passing library for the Paragon, and the MPLp message passing library for the SP1. This paper describes the issues involved in porting this library to different message layers, and it shows that the library can be implemented efficiently on these message layers. This paper presents microbenchmarks and performance of applications for all the implementations. It also discusses the primitives exported by the library and the library's relationship to languages.


BibTeX citation:

@techreport{Luna:CSD-94-810,
    Author = {Luna, Stephen S.},
    Title = {Implementing an Efficient Portable Global Memory Layer on Distributed Memory Multiprocessors},
    Institution = {EECS Department, University of California, Berkeley},
    Year = {1994},
    Month = {May},
    URL = {http://www.eecs.berkeley.edu/Pubs/TechRpts/1994/5605.html},
    Number = {UCB/CSD-94-810},
    Abstract = {Libsplit-C is a library of communication and global storage primitives for distributed memory multiprocessors. The library has been implemented on the CMAM and CMAML active message layers for the CM5, the HPAM active message layer for an FDDI network of HP workstations, the NX message passing library for the Paragon, and the MPLp message passing library for the SP1. This paper describes the issues involved in porting this library to different message layers, and it shows that the library can be implemented efficiently on these message layers. This paper presents microbenchmarks and performance of applications for all the implementations. It also discusses the primitives exported by the library and the library's relationship to languages.}
}

EndNote citation:

%0 Report
%A Luna, Stephen S.
%T Implementing an Efficient Portable Global Memory Layer on Distributed Memory Multiprocessors
%I EECS Department, University of California, Berkeley
%D 1994
%@ UCB/CSD-94-810
%U http://www.eecs.berkeley.edu/Pubs/TechRpts/1994/5605.html
%F Luna:CSD-94-810