You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Garrett Rooney <ro...@electricjellyfish.net> on 2006/04/26 19:35:53 UTC
Re: svn commit: r19456 - trunk/subversion/libsvn_fs_fs
On 4/26/06, lundblad@tigris.org <lu...@tigris.org> wrote:
> Author: lundblad
> Date: Wed Apr 26 12:30:40 2006
> New Revision: 19456
>
> Modified:
> trunk/subversion/libsvn_fs_fs/dag.c
> trunk/subversion/libsvn_fs_fs/dag.h
> trunk/subversion/libsvn_fs_fs/fs_fs.c
> trunk/subversion/libsvn_fs_fs/fs_fs.h
> trunk/subversion/libsvn_fs_fs/tree.c
>
> Log:
> In FSFS, decrease the server load somewhat in a common case where
> the filesystem has actually stored the delta the client is asking for.
Nice! Does this give us a noticable speedup in any benchmarks?
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r19456 - trunk/subversion/libsvn_fs_fs
Posted by Garrett Rooney <ro...@electricjellyfish.net>.
On 4/26/06, Peter N. Lundblad <pe...@famlundblad.se> wrote:
> I tried blame locally and it gave maybe 30% on that particular test. Replay
> seemed *somewhat* faster, but it didn't save the world or
> something.. Take these numbers with a truck full of salt, but still,
> just tossing out a delta for every other file against its immediate
> predecessor justified this change. There are more opportunities: just
> combining deltas when the source is a predecessor to some previous
> version of the target and avoiding one of the delta applications when
> the two file have a common ancestor. That can be done later.
Excellent, thanks.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r19456 - trunk/subversion/libsvn_fs_fs
Posted by "Peter N. Lundblad" <pe...@famlundblad.se>.
Garrett Rooney writes:
> On 4/26/06, lundblad@tigris.org <lu...@tigris.org> wrote:
> > Author: lundblad
> > Date: Wed Apr 26 12:30:40 2006
> > New Revision: 19456
> >
> > Modified:
> > trunk/subversion/libsvn_fs_fs/dag.c
> > trunk/subversion/libsvn_fs_fs/dag.h
> > trunk/subversion/libsvn_fs_fs/fs_fs.c
> > trunk/subversion/libsvn_fs_fs/fs_fs.h
> > trunk/subversion/libsvn_fs_fs/tree.c
> >
> > Log:
> > In FSFS, decrease the server load somewhat in a common case where
> > the filesystem has actually stored the delta the client is asking for.
>
> Nice! Does this give us a noticable speedup in any benchmarks?
>
I tried blame locally and it gave maybe 30% on that particular test. Replay
seemed *somewhat* faster, but it didn't save the world or
something.. Take these numbers with a truck full of salt, but still,
just tossing out a delta for every other file against its immediate
predecessor justified this change. There are more opportunities: just
combining deltas when the source is a predecessor to some previous
version of the target and avoiding one of the delta applications when
the two file have a common ancestor. That can be done later.
Regards,
//Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org