[PATCH 5 of 7 v2] bdiff: give slight preference to longest matches in the middle of the B side

[Gregory Szorc]

On Tue, Nov 15, 2016 at 1:13 PM, Mads Kiilerich <mads at kiilerich.com> wrote:

> On 11/15/2016 09:57 PM, Mads Kiilerich wrote:
>
>> # HG changeset patch
>> # User Mads Kiilerich <madski at unity3d.com>
>> # Date 1478626653 -3600
>> #      Tue Nov 08 18:37:33 2016 +0100
>> # Node ID 5ae7e8061a9671e8941a7a17e316254d228acf59
>> # Parent  5bb26f29b1509520ca3af4c540775cab50b4d6c0
>> bdiff: give slight preference to longest matches in the middle of the B
>> side
>>
>
> This and the following patches can probably be rearranged and folded to
> give less churn. I would however appreciate to get another round of
> thorough review with this structure - that seems to me to give a more
> natural progression.
>
> The benefit from these changes is mainly "better diffs". Better diffs do
> not necessarily compress better and there might be some small increases in
> actual diff size. I have not noticed any significant performance changes.
>
> Greg, can you verify it doesn't impact your bdiff benchmarks in your
> environment in a bad way?
>

I /think/ my perfbdiff changes have all mostly landed. So you should be
able to perform benchmarks. I've been using
https://hg.mozilla.org/mozilla-unified for many measurements.

If you want me to double verify, could you please push this series to a
repository somewhere? I don't have Mercurial configured to apply patches
from the mailing list and don't want to go through the hassle of doing that
:(
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.mercurial-scm.org/pipermail/mercurial-devel/attachments/20161115/968cd57a/attachment.html>