discussion forum


message

Name: james
Date: June 16, 2003 at 12:29:35
Subject: Re: Hyperthreading and 'lock' prefix
In Reply To: Re: Hyperthreading and 'lock' prefix by
Stephan Keil on June 16, 2003 at 06:07:58
Text: | In general you are right. But with my special problem I
| incrementing/decrementing some reference counts, which is done by
| processors only. Unfortunately these reference counts are
| accessed frequently and I cannot afford the performance penalty.

Can you afford that your program may or may not work reliably on
a future HTT-capable processor? I mean, the P4 might be capable
of detecting this particular scenario, and as a result work fine
even if you don't use the LOCK prefix. However, architecturally
the LOCK is required, to ensure compatibility -- no ifs or buts.

Performance isn't that good an excuse for writing implementation-
dependent code. Also, bundling more than one version of the code
and selecting the desired variant at run-time requires that you
come up with an absolutely reliable processor/feature detection.
Which you won't be able to, unless you can foresee the future.

Are you absolutely certain that you can't improve the algorithm,
to avoid this particular performance issue? Have you considered
getting help from Intel's developer support team? Just curious.


post a followup message
(Be nice... or be blocked. Be technical... or be erased.)

Name: optional link title:
eMail: optional link URL:
Subject: optional image URL:
  Insert line breaks by hand when only about one inch remains at the right side. Otherwise your message will be unreadable.
Text:
 

  Note: The above eMail form fields may look unaligned if you are using a browser other than Netscape Navigator version 3.0.


currently posted followup messages
(You may have to press the RELOAD button of your browser.)



main page