Subj : Re: pthread priority in Linux 2.4 Kernel
To   : comp.programming.threads
From : David Schwartz
Date : Tue Aug 02 2005 08:31 pm


"David Hopwood" <david.nospam.hopwood@blueyonder.co.uk> wrote in message 
news:izTHe.70989$dN6.53465@fe1.news.blueyonder.co.uk...

> It doesn't even mean that. This is true, approximately (for both the old 
> and
> the O(1) Linux schedulers), if one or both of the threads are "realtime" 
> threads;
> otherwise what determines which thread is chosen is the dynamically 
> computed
> 'goodness' value, which may be quite different from the static priority.

    I wasn't quite clear. By "priority", I mean static priority among 
FIFO-scheduled threads. Normally, priority is, as you say, modified 
dynamically.

    In any event, you *cannot* use priority as an exclusion technique. Some 
unexpected problem can always block a thread, such as a page fault, and 
allow the "wrong" thread to run. You *MUST* engineer things with 
synchronization techniques so that there either is no wrong thread or the 
wrong thread *cannot* run.

    DS

.