TGM

Task Graph Manager

Short Description

Support for structuring interactive programs as concurrent tasks (threads) connected in a prioritized dataflow graph that conveys multi-resolution results (dissertation work).

Contact

George Forman
Computer Science & Engineering
University of Washington
Seattle
USA

email: forman@cs.washington.edu

World Wide Web

George Forman

Description

In modern computing environments, shared resources (e.g. remote web servers and wireless networks) deliver highly variable service rates. Structuring interactive programs as concurrent tasks (threads) that produce a series of increasing resolution/quality results gives the application-independent runtime system the ability to adjust resource scheduling (priorities) to give the user good perceived response time for the items they are focussed on at the moment.

The implementation consists of a C++ pre-processor to convert language extensions into multi-threaded code and runtime support for managing task graphs-- priorities, cancellation, multi-resolution results, synchronization...

The LEDA library was quick to use and has been reliable.