db_connect: Could not connect to paper db at "wotug@dragon.kent.ac.uk"<br>
db_connect: Could not connect to paper db at "wotug@dragon.kent.ac.uk"<br>

%T An efficient global convergence detection scheme for parallel algorithms on transputer network
db_connect: Could not connect to paper db at "wotug@dragon.kent.ac.uk"<br>
%A  K. G. Kumar,  A. Basu,  S. Srinivas, A. Paulraj
db_connect: Could not connect to paper db at "wotug@dragon.kent.ac.uk"<br>
%E  Stephen J. Turner
%B OUG\-12: Tools and Techniques for Transputer Applications
%X The paper discusses a novel and efficient scheme for the
   detection of global convergence in parallel iterative
   algorithms. Convergence information is maintained as
   processes which execute concurrently with computations of
   the algorithm and exploit the asynchrony inherent in the
   progress of most executions of such algorithms towards
   global convergence. The scheme treats messages signalling
   convergence as having a lower priority as compared to those
   signalling non\-convergence. It minimizes the waiting time
   at the end of iterations for convergence related
   communications. Analytical results indicate that the global
   convergence detection scheme proposed in this paper is
   faster than the methods proposed earlier [3]. The scheme is
   particularly suited for implementation on Transputer based
   parallel machines. An OCCAM implementation of the scheme on
   a torus of Transputers is described together with a method
   for experimental verification of the analytical results.