Date: Wed, 31 Mar 1999 02:56:48 +0000
Reply-To: John Whittington <medisci@POWERNET.COM>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: John Whittington <medisci@POWERNET.COM>
Subject: Re: Solution times of problem across SAS versions and platforms
Content-Type: text/plain; charset="us-ascii"
At 17:51 30/03/99 -0600, Gullion Christina wrote:
>I tried doing one of these time experiments not too long ago and discovered
>the job took a different amount of time everytime I ran it. Trivial things
>like moving the mouse would change the time. What other programs were
>running. If any "automated" events occurred, like a file save or a registry
>update, or receipt of an e-mail message. As I was reminded, this kind of
>thing requires say 1000 replications, then look at the mean or median.
Chris, that's very true - and is one reason why one cannot, unfortunately,
'scale down' these tests too much and expect to get reliable answers.
However, once one is into exceution times measured in tens of minutes or
more, at least on a PC that is not doing anything else 'of any significance'
at the same time, one generally gets pretty consistent answers over re-runs.
There are certainly some exceptions - for example, the first step(s) in a
SAS session always seem to run far more slowly - perhaps because the
compiler has to be loaded. One can also get some very misleading figures
because of the effect of caches - I/O operations may be very much faster if
one immediately re-runs the same code.
Kind Regards
John
----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: medisci@powernet.com
Buckingham MK18 4EL, UK mediscience@compuserve.com
----------------------------------------------------------------
| 
