On compressing revlogs

[Bryan O'Sullivan]

On Sun, Jun 24, 2012 at 8:53 AM, Na'Tosha Bard <natosha at unity3d.com> wrote:

>
> Just curious if there are plans to move this forward or not?  We're also
> getting hit by horrible push performance when we push a large number of
> changesets with the vast majority of time spent in zlib.decompress() while
> the CPU is maxed out.
>

I decided that trying to cache manifest revisions was not productive. It's
easy to verify that if the write side does its job well, the read side
absolutely helps performance, but it's very hard to figure out a sensible
strategy for the write side.

Instead, I've been focusing on an extension that lets Mercurial optionally
use lz4 compression for revlogs. This is very nearly as fast as caching a
manifest, but it works reliably, and for all revlogs, and speeds up all
kinds of operations very nicely. At a cost, of course: .hg is about 25-30%
bigger than when using zlib revlogs.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurial-scm.org/pipermail/mercurial-devel/attachments/20120625/671320ad/attachment-0002.html>