Search

Parallel Computing Parallel Computing

PREMA

The Load Balancing Layer (LBL) provides support for dynamic load balancing; LBL has been implemented on top of DMCS/MOL and CLAM.

MOL

Description

No Longer supported

Application Program Interface

The functionality provided by MOL can be broken into three broad categories.

  • The first group is made up of Environment functions, and are used to initialize and shutdown MOL in an orderly fashion.
  • The second group of functions are message passing functions like:
    • mol_message - copy a data buffer to a remote processor
    • mol_request - copy a data buffer to a remote node and execute a user define handler
  • The third group is the mobile object and pointer creation and maintainance group. For more details see please see the manpages.

Platforms

  • Unix
    • Linux + MPI
    • Solaris + MPI
    • AIX + LAPI
  • Microsoft Windows
    • Windows 2000 + MPIPro

Performance Data

Documents

Man Pages

Papers

Mobile Object Layer: A Runtime Substrate for Parallel Adaptive and Irregular Computations, 
Nikos Chrisochoides, Kevin Barker, D&eactue;mian Nave, and Chris Hawblitzel, published in Advances in Engineering Software. 
[HTML] [GZIPPED POSTSCRIPT, 113Kb]

User Guide

(coming soon)

Installation Guide

(coming soon)

  • Unix Platforms
  • Microsoft Windows

Downloads

Please contact Nikos Chrisochoides

 

Applications & Users

Applications

  • Parallel 3D Guaranteed Quality Delaunay Triangulation
  • Parallel 2D Constrained Delaunay Triangulation
  • Prototype of Multi-Layer Runtime System for Petaflops Architectures
  • Parallel Quadtree Data Structure

Users

  • Cornell University
  • Mississippi State University
  • University of Notre Dame

Contact Info

Andriy Kot, Phone: (757) 221-3469
Computer Science Department FAX: (757) 221-1717
Parallel Experimental Systems Lab URL: http://www.cs.wm.edu/~kot
College of William and Mary Email: This email address is being protected from spambots. You need JavaScript enabled to view it.
Williamsburg, VA 23187-8795 P.O. Box 8795

 

Nikos P. Chrisochoides, Phone: (757) 221-3466
Computer Science Department FAX: (757) 221-1717
Parallel Experimental Systems Lab URL: http://www.cs.wm.edu/~nikos
College of William and Mary Email: This email address is being protected from spambots. You need JavaScript enabled to view it.
Williamsburg, VA 23187-8795 P.O. Box 8795

Development Team

Professors

Graduate Students

DMCS

Description

No Longer supported

Application Program Interface

The functionality provided by DMCS can be broken into three categories.

  • Environmental: The first group of functions are common to every parallel programmer. This group includes initialization and shutdown routines as well as functions to query the system about the number of processors running as well as the calling processor's ID. For more details please see themanpages.
  • Remote Memory Manipulation: The second group of functions are use to access remote memory. Operations to allocate and free remote memory are provided as well as operations to read and write to remote memory. The read/write operations are available in both synchronous and asynchronous versions.
  • Remote Service Requests (RSRs): These are more commonly known as Remote Procedure Calls (RPCs) and Remote Method Invocations (RMIs). These are used to invoke user-level functions on remote processors. The user-level handlers are able to take between 0 and 4 arguments. Another variety of RSRs, called Put-Ops, are able to receive a user-level buffer of arbitrary size. These methods are available in synchronous and asynchronous flavors, and are unable to propagate a return value.

PERFORMANCE DATA

The following graphs depict results from two tests we have run. The tests consist of timing the dmcs_async_put_op( ) on two different architectures. The first was run on a cluster of PIII 1Ghz machines running Linux connected by 100 Mb fast-ethernet. The second was run on a cluster of Sun Ultra 5 333 Mhz machines connected by fast-ethernet. [VIEW GRAPHS]

Documents

Man Pages

Papers

  • Data Movement and Control Substrate for Parallel Adaptive Applications,
    Kevin Barker, Nikos Chrisochoides, Jeffrey Dobbelaere, Démian Nave, and Keshav Pingali
    [HTML] [POSTSCRIPT, 420 kb]

User Guide

Coming soon

Installation Guide

Coming soon

  • Unix Platforms
  • Microsoft Windows

Downloads

DMCS Source Code

[dmcs-1.0.tgz]

  • DMCS for UNIX compatible systems
    • An MPI implementation
    • A networked UNIX-like system
    • DMCS for Windows 2000/NT systems (coming soon)

Applications & Users

Applications

  • Parallel 3D Guaranteed Quality Delaunay Triangulation
  • Parallel 2D Constrained Delaunay Triangulation
  • Prototype of Multi-Layer Runtime System for Petaflops Architectures
  • Parallel Quadtree Data Structure

Users

  • Cornell University
  • Mississippi State University
  • University of Notre Dame

Contact Info

Andriy Kot, Phone: (757) 221-3469
Computer Science Department FAX: (757) 221-1717
Parallel Experimental Systems Lab URL: http://www.cs.wm.edu/~kot
College of William and Mary Email: This email address is being protected from spambots. You need JavaScript enabled to view it.
Williamsburg, VA 23187-8795 P.O. Box 8795

 

Nikos P. Chrisochoides, Phone: (757) 221-3466
Computer Science Department FAX: (757) 221-1717
Parallel Experimental Systems Lab URL: http://www.cs.wm.edu/~nikos
College of William and Mary Email: This email address is being protected from spambots. You need JavaScript enabled to view it.
Williamsburg, VA 23187-8795 P.O. Box 8795

Development Team

Professors

CrayDMCS

The CrayDMCS library implements DMCS on top of Remote Memory Access Communication subsystem for the Cray XT3 architecture

CLAM

The Communication Layer for Asynchronous Mobile (CLAM) computations is designed to combine the functionality of DMCS and MOL in order to improve efficiency and incorporate the lessons we learned from the use of its predecessor systems