[lkml]   [1998]   [Sep]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: slowdown 2.0.35 vs 2.1.121 (pre 1)
On 10 Sep 1998, Claude Gamache wrote:

> Hi Simon
> > built 2.1.121 while running 2.0.34 and then built it again (after doing make
> > clean; make xconfig; make dep) on top of the 2.1.121 previously built
> >
> > there was a 6% slowdown 9:14 vs 8:51
> > is this to be expected?
> We observed similar performance loss with kernels
> I also saw a few posts regarding slowdown of the Xserver. The bottom
> line is not only gcc or the Xserver are slowing down, but all the
> processes are slowing down. The problem (performance loss or if you
> wish, the slowdown) occurs after the system has been heavilty loaded
> (building the kernel is an example of an heavy load for the system)

I tried your (below) test repeatedly, and could not replicate your results.
I ran 50-70 additional instances (P150-MMX) after tuning for 15 secs clean,
and always came back to pristine condition.

> > i know the test was a bit rough and ready
> > is 6% statistically significant?

In this case, he disabled dma.. reading the source tree could account
for his performance loss.

> We observed worse performance loss under certain conditions. So I
> believe that your measurement is quite significative.

In all tests I've run recently, is usually slightly faster than
2.0.xx when it comes to compiling. The only case I've seen where it isn't
is where some very strange dcache (bug?) effect kicks in. I posted some
_strange_ profiling results wrt this a while back.. and have not had it
pop up for a few versions. (Trying to explore it led to only one result..
doing a umount/mount of the affected fs cured the situation.)


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:44    [W:0.046 / U:2.896 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site