Global Arrays

Abstract

The Global Array (GA) toolkit provides an efficient and portable shared-memory programming interface for distributed-memory computers. Each process in a MIMD parallel program can asynchronously access logical blocks of physically distributed matrices, without need for explicit cooperation by other processes. Unlike other shared-memory environments, the GA model exposes to the programmer the Non-Uniform Memory Access (NUMA) characteristics of the High Performance Computers and acknowledges that access to remote data is slower than to local data.

DateOfInformation

Fri Oct 30 13:08:54 1998

Domain

Parallel Processing Tools!Parallel Programming Libraries

Keyword

virtual shared memory; MIMD machine; nonuniform memory access; distributed array; execution trace visualization; parallel programming library

Name

Global Arrays

TargetEnvironment

Distributed-memory MPP: IBM SP-1, SP-2, SP ; CRAY T3D and T3E; Intel IPSC/860, Delta, Paragon; Fujitsu VPP-700, VX-4; NEC Cenju Shared-memory systems: KSR-1/2; CONVEX SPP-1200; SGI PowerChallenge and Origin-2000 Networks of workstations (including clusters of SMPs): SUN (SunOS and Solaris); SGI (32- and 64-bit); IBM RS/6000 ; DEC Alpha ; PC under Linux ; PC under Windows NT with Fast Messages (HPVM Project, University of Illinois) In addition, research/prototype implementations have been developed for: Intel Paragon running Sunmos with Active Messages; metacomputing environments

Version

2.4

Webpage

http://www.emsl.pnl.gov:2080/docs/global/ga.html

ContactIs

Jaroslaw Nieplocha