lkml.org 
[lkml]   [1998]   [Sep]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRE: 2.1.123 (Possibly slower)
On Tue, 29 Sep 1998, Richard B. Johnson wrote:

> On Mon, 28 Sep 1998, Adam D. Bradley wrote:
>
> > mean([322,355,362,328,311,306,329,341,332]) = 331.7778
> > var([322,355,362,328,311,306,329,341,332]) = 343.9444
> > std([322,355,362,328,311,306,329,341,332]) = 18.5457
> >
> > mean([416,408,398,328,422,432,381,396,398]) = 397.6667
> > var([416,408,398,328,422,432,381,396,398]) = 918.5000
> > std([416,408,398,328,422,432,381,396,398]) = 30.3068
> >
> > Interesting that the increase in standard deviation is disproportionate
> > with the increase of the mean... hmm...
>
> Object files are not likely to be put the same place on the disk
> because they are all deleted ahead of time, therefore I would expect
> that the standard deviations would not track since a major bottleneck
> is I/O. About all one could expect is that the kernel would build
> at _about_ the same rate.

I'm not quite clear on your meaning here. Are you perhaps thinking of
"filesystem rusting" effects? I would expect (although I could very well
be mistaken) that variability caused by the filesystem and sub-filesystem
stuff (buffer cache and disks) _WOULD_ appear, in proportion, in both sets
of data, since (presumably) the filesystem has already "matured" and so
the cost of allocating new files will consistently be drawn from a fairly
stable distribution. Perhaps the greater variability is the _result_ of a
change in the I/O subsystem which causes I/O to be a more significant
bottleneck? (Did the driver for the disk this filesystem resides on
change between the kernel versions?)

Another question: did you have a "MAKE=make -j2" line (or similar) in the
top-level Makefile (i.e., were these parallel builds), and what was the
degree of parallelism? (A profile of the load averages during the builds
might be interesting...)

Adam
--
You crucify all honesty \\Adam D. Bradley artdodge@cs.bu.edu
No signs you see do you believe \\Boston University Computer Science
And all your words just twist and turn\\ Grad Student and Linux Hacker
Reviving just to crash and burn \\ <><
---------> Why can't you listen as love screams everywhere? <--------


-
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    [W:2.648 / U:0.012 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site