lkml.org 
[lkml]   [1998]   [Sep]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Interesting scheduling times - NOT
Larry McVoy wrote:
>
> Just to put to rest the idea that maybe the median is covering things
> up, here's the full set of data, note the small standard deviation:
>
> 2 7.58 (7.74 7.65 7.63 7.60 7.60 7.58 7.58 7.58 7.55 7.54 7.53)
>
> Here's the same thing with each run taking 500 milliseconds (so a total of
> about 6 seconds of run time):
>
> 2 7.73 (8.38 8.14 8.13 8.00 7.94 7.73 7.62 7.60 7.46 7.33 7.04)
>

Just to scroll things back a few days here guys: (I just noticed this
message in my backlog)

Anybody else notice that Larry's results are monotonic on both sets of
data ? The chances against monotonicity for each run are 2^9 to 1
against.

This says VERY loudly that these are not random differences - something
else is happening. That something else appears to dwarf the random
effects that are probably present, because monotonicity isn't violated
even a single time.

Then compare first and last entries for the second run and you get a
non-trivial difference...

The fact that the times are varying even until the end of each run says
that something hasn't equilibrated (probably a cache thing) and that
drawing conclusions from the results is subject to some dangers.

And yeah, the median certainly covered *that* thing up, no?

Comments?

Neil

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:44    [from the cache]
©2003-2011 Jasper Spaans