Re: Stablilizing execution time?

Mike Black (mblack@csihq.com)
Sun, 21 Jul 1996 07:16:48 -0400


At 07:48 AM 7/19/96 +0300, Linus Torvalds wrote:
>
>On Thu, 18 Jul 1996, Hubert A. Bahr wrote:
>>
>> What can be done beyond running in single user mode to
>> bring execution time to a more repeatable state.
>
>You are unlikely to get much system time variation, as the kernel seldom
>does anything which is very susceptible to cache effects. Most of the
>time difference is likely to be in user mode or in IO effects (and I
>assume you have minimized those effects? I assume it's CPU-bound?)
>

Unfortunately Hubert doesn't who his times...sigh...

I expect that his major problem is between the first and second execution.
The first execution should always be slower that the second as it will
rearrange the cache for it's own use. This can be quite noticeable if
Hubert is running X with 16meg as it will have to start swapping in order to
accomodate 4Meg of data. 2nd and subsequent executions should show fairly
consistent behaviour as long as NOTHING else is done between them to change
the cache arrangement.

This does bring up an interesting point though...is there (or should there
be) a cache flush program? Would be real handy in situations like this.
Something which invalidates all cache pages would allow for consistent bench
marking.

-
/----------------------------------------------------------\
| Mike Black mblack@csihq.com 407-676-5118, x203 |
| Computer Science Innovations, Inc. 407-676-2355 FAX |
| Melbourne FL 32904-2314 http://www.csihq.com |
\----------------------------------------------------------/