Discovery of Computation Structure

Introduction

The programs that developers write specify most, but not all, of the computation components that run when an application program executes. Often, these additional utility components at the user and system level support the executing program in ways that are of no concern to the developer. However, under some application semantics, precisely producing the desired behaviour requires that all computation components be explicitly known and controlled. Discovery is a particular use of the KUSP CCSM, DSKI data gathering, and DSKI Active Filtering capabilities which enables the system to discover the complete set of components required to fully execute each computation. The computation structure information can be used within the OS at run-time by Group Scheduling to provide customized execution control, by user-level tools at run-time to monitor what is occurring within the system for any set of computations on the system, and in a post-processing mode to fully analyze, and in some cases visualize, many aspects of computation structure and behavior.

The system programmer is free to use this information about computation structure in any way that seems appropriate. In some cases, this appropriate action consists merely of recording all components uses for evaluation purposes. For example, an embedded system developer may wish to know the set of all computation components on the system used by the specified list of applications as an aid to removing any unnecessary components to minimize footprint and power consumption. In another case the appropriate action may consist of adding discovered computation components to a group representing the computation in order to produce desired computation behaviour. For example, a graphic based application requiring precise control over display of the produced graphics would first discover its connection to the X server and add the X server to its scheduling group to ensure prompt display

Contents:

Indices and tables

Table Of Contents

Next topic

Use of Data Streams in Discovery

This Page