From: Colin Nevin (colin_nevin at domain yahoo.com)
Date: Fri 06 Sep 2002 - 11:59:51 IST
> Think that's bad! Eircell bought a mad
> mainframe class machine from Compaq (GS160) to
> run their billing app faster. But it ran at exactly
> the same speed for some reason. I was called in
> and promptly found a sleep(1) synchronisation
> between 2 processes! So much for throwing hardware
> at a problem.
So much for throwing hardware at a problem indeed, I
have heard of some unscrupulous contractors doing that
so they are called back in to 'optimize' the
As my own problem as the parent process controls all
i/o so just drawing a screen (using curses), involves
hundreds of semop calls to syncronise with the child
so adding a delay no matter how small will have a
detrimental effect on preformance; I have see examples
of using semop that have a sleep between semop calls
to 'allow time for the semaphore to change state' ??
Surely this is not necessary or am I missing
Perhaps there are some compile options I can try, or
can I use a different implemenation of semaphores if
available for Linux.
thanks for the pointers though,
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
This archive was generated by hypermail 2.1.6 : Thu 06 Feb 2003 - 13:18:42 GMT