slndstrm@SIERRA.STANFORD.EDU (Stephen F. Lundstrom) (07/22/89)
Here, at Stanford, we are investigating how to do dynamic processor
allocation for networks of communicating processes. Since processor
allocation is done at run-time, our technique is able to address a wider
class of applications. For applications that define networks of
communicating processes during their course of program execution, the
conventional approach to preallocate processors fails - it is either
infeasible or ineffective. In an environment where the number of processors
available to the application varies from time to time, static processor
allocation is also inappropriate. Our approach is able to circumvent these
difficulties and targets for sub-optimal allocation and scheduling. We
have already implemented a run-time system to support our approach. Now
we badly need a variety of test cases to evaluate the method.
Do you have any application program that might be a good candidate to try
out the above method? It will be perfect if they are written in C and in
a programming paradigm close to CSP (i.e. during program execution, a
network of communicating sequential processes is defined and then
subsequently created to carry out certain subtask). If so, we would like
to hear from you.
If you are developing programs for applications that require dynamic
processor allocation as stated above, we would also like to get in touch
with you.
Thanks for your attention.
------------
You can contact us as follows:
ngai@umunhum.stanford.edu
(Tin-Fook Ngai)
slndstrm@sierra.stanford.edu
(Prof. Stephen F. Lundstrom)
Computer Systems Lab, ERL 455
Stanford University
Stanford, CA 94305
(415) 723-0140
------------