Measurement of timings

[John Campbell <J.Campbell@cs.ucl.ac.uk>]

Does anybody out there have practical experience of obtaining accurate and
reliable timings for execution of functions, where these timings are values
available for subsequent use inside programs? We have been doing the obvious
things, but the values that we get seem to be too coarse-grained and
subject to surprising variations. 

I'll be grateful if any comments can be sent by email both to me
(jac@cs.ucl.ac.uk) and to N.Chatterjee@cs.ucl.ac.uk.

John Campbell