Request for review of the svnsync_ra_serf branch.

So, I finished the performance improvements for svnsync on the
svnsync_ra_serf branch, and would like a review before I merge it back
to trunk.

There's an svnsync part and a ra_serf part to this code, which can be
reviewed independently, as the new svnsync code also works with all ra
layers other than ra_serf.

The way 'svnsync synchronize' is now implemented is explained in a
previous email on this topic:
http://svn.haxx.se/dev/archive-2007-11/0223.shtml

I did some testing with the new code and the performance gains are
significant. Some numbers:

* First 200 revisions of http://svn.collab.net/repos/svn to file://

trunk trunk branch branch
neon serf neon serf
-------------------------
2m38s 1m56s 2m30s 0m34s
100% 73% 94% 22%

* First 200 revisions of https://serf.googlecode.com/svn to file://
  (these are mostly empty revs)

trunk trunk branch branch
neon serf neon serf
-------------------------
4m20s 4m18s 4m13s 1m39s
100% 99% 97% 38%

I have no accurate data on memory usage, but it seems to be only a bit
higher than trunk svnsync and slightly increasing after 100's of revs.

Lieven

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Nov 8 16:47:25 2007