Introduction

As you have read probably before, the inconvenient term DP(HIGH) should indicate, that it stands for a set of functions which are on one side actually only one part of the DP Toolbox Set, but on the other side these functions can be seen as the real core of the entire toolbox set. In that sense and for better readability we will use the short term DP-Toolbox instead of DP(HIGH) throughout this user's guide.

The DP-Toolbox is realized on top of the DPLOW-Toolbox and supports a very simple model for parallel and distributed computing in MATLAB. The model is based on autonomous DP instances. Autonomous means that all actions and states inside a DP instance are caused exclusively by the instruction stream which is provided to the interpreter via the standard input (keyboard in case of an interactive session or file in batch mode). The only difference between a DP instance and an ordinary MATLAB instances is, that a DP instance is an enrolled process at the PVM system. With simple words: DP instances are able to communicate with the outside world - ordinary MATLAB instances not.

Beside the ability to communicate, DP instances can create new instances. Between the ``spawner'' and the spawned instance there exist a parent/child relation. Over and above that there are no further ordering relations in a set of DP instances. Therefore, no central or distributed control instance is known in the DP Model.

The two major objectives of the DP model are:

• bringing up the concept of message passing to a level that goes together with MATLAB (array passing), and
• hiding all the snares that occur when MATLAB processes (especially interactive ones) have to be launched.

The limits of the DP Model are:

• Due to the autonomy of DP instances only instructions that come in over the standard input stream can be processed. Therefore, the remote execution of MATLAB programs (M-scripts, M/MEX-functions, built-in functions) is not possible.
• The opportunity to pass a MATLAB program during starting up a MATLAB instance that executes the program immediately is only suitable for large applications in the production phase. (because of the long startup times of MATLAB processes in the range of several seconds)