[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: performance on cygwin

From: Matthew Bentham <mjb67_at_artvps.com>
Date: Wed, 10 Feb 2010 12:47:06 +0000

stsp_at_stsp.name wrote:
> On Wed, Feb 10, 2010 at 11:57:35AM +0000, Matthew Bentham wrote:
>> I've been hearing rumours of performance improvements to Windows svn
>> clients with the 'wc-ng' work. A couple of weeks ago I decided to
>> have a go with subversion trunk to see what it would make of our
>> repository. I've compiled it, compared the performance with
>> subversion 1.6 from cygwin, and done some profiling with Intel
>> VTune. The results follow, I hope they are interesting.
>
> Thanks for profiling the code. This is quite interesting and
> I think the results are in line with our expectations.
>
> We're aware that trunk is currently slower than 1.6.x.
> The major reason for the current performance problems in trunk is that
> the .svn directories have not been centralised yet. Once there is only
> a single .svn directory at the root of the working copy, we expect a
> large speed-up since only a single sqlite database will be used.
> It would be great if you could profile the code again once we have
> reached that goal. If you want to help us reach that goal faster,
> please join the wc-ng development efffort :)

OK! I'll track changes for now, and I'm here if anyone wants me to try
out something specific.

> The current non-performance of trunk has been discussed before,
> e.g. see http://svn.haxx.se/dev/archive-2009-09/0189.shtml
> and related messages.

Thanks, somehow I missed that thread when reading the archives before
posting.

Just for fun I tried building with "PRAGMA journal_mode=MEMORY", the
result is:

$ time /cygdrive/e/subversions/subversion/profinstalldir/bin/svn.exe update
At revision 36896.

real 0m17.685s
user 0m6.000s
sys 0m7.109s

 From my original post, running subversion 1.6:

> $ time svn update
> At revision 36888.
>
> real 0m19.865s
> user 0m0.250s
> sys 0m1.468s
> $

Which I think is pretty conclusive evidence (if any more were needed)
that the performance problems will be solved by reducing the database IO
operations.

All the best,
Matthew
ART VPS
Received on 2010-02-10 13:47:46 CET

This is an archived mail posted to the Subversion Dev mailing list.