is an infrastructure for the creation of Problem Solving Environments. Specifically, this project grew out of the need to solve a number of different classes of engineering and science problems on a variety of high-performance computer architectures. CECAAD attempts to identify all the common elements of problem solving environments, then provide some support for creating those elements. At present, CECAAD includes the following elements:
ADF -- The Algorithm Description Format
ADF is a generic format for storing an attributed graph. Virtually any algorithm in any application can be stored as an attributed graph one way or another. The attribute format makes ADF extremely flexible -- not only can it represent many different problems, it can represent them at many different levels of abstraction. The components of ADF are implemented as a set of Java classes. The public interface to them can be found here or by following the link below in the documents section.
Agent Model -- Agents are entities which manipulate ADF designs. CECAAD provides a number of agents, and extensive support for creating new agents.
ADF Manager-- The ADF Manager is responsible for coordinating the actions of the agents on the shared design. The primary role of the ADF Manager is to handle all I/O associated with ADF designs. The ADF Manager also manages libraries of ADF designs. Libraries exist either as SQL databases, or as directories on a filesystem. The ADF Manager supports a rich set of queries on the design libraries, but only a subset of the searching functionality works if the library is not in a database. The Manager allows agents (and therefore users) access to both types of libraries simultaneously. The manager also keeps all running agents notified of changes in the currently active designs. The programmers interface to the ADF Manager, as well as more information about its internal structure, can be found here .
Displays -- Displays are windows where agents can interact with users. Multiple agents can interact with a user through a single display, and agents can move between displays or use multiple displays. An agent describes to the display how it wishes to appear and interact with the users through its views. More details about displays and views can be found in the Agent Writer's Guide
Common Shared Agents:
Task Partitioning Tool
Problem Solving Environment Infrastructure for HPC Systems
Brian's Master's thesis on CECAAD
Interface to the ADFmanager
Agent interface to ADF designs
Poster Presentation for SuperComputing '98
SPIE paper on RCADE
Aerospace paper on RCADE
Dr. Walt Ligon
is the faculty advisor and general overlord of all the PARL lab activities.
Dan Stanzione is the PhD student heading the CECAAD and Parallel Programming Environment projects, and is also involved in the RCADE project. Among his other duties are running this web page, system administration of most of the PARL lab, and teaching a number of undergraduate courses. Dan is also responsible for all the silly acronyms on these pages. If you have complaints or comments, you probably want to contact Dan (see contact info above).
Brian Boysen is an MS student just completing his degree and going off to work for Lucent Technologies. He did a lot of the actual coding on the CECAAD infrastructure, including all the work on the ADF Editor. Click on the link above in documents to check out his thesis
The Parallel Architecture Research Lab
Clemson University's Department of Electrical and Computer Engineering
Problem-Solving Enviroment Related Links:
Problem Solving Environments Research Group
EPOCH Home Page
Parallel Processing Related Links:
These two links provide a pretty extensive listing of parallel processing stuff, so I won't repeat it all hear