[PATCH 3 of 3] revlog: optimize _chunkraw when startrev==endrev

[Gregory Szorc]

On Sun, Nov 6, 2016 at 7:16 AM, Pierre-Yves David <
pierre-yves.david at ens-lyon.org> wrote:

>
>
> On 11/04/2016 09:27 AM, Yuya Nishihara wrote:
>
>> On Tue, 01 Nov 2016 18:25:47 -0700, Gregory Szorc wrote:
>>
>>> # HG changeset patch
>>> # User Gregory Szorc <gregory.szorc at gmail.com>
>>> # Date 1477244433 25200
>>> #      Sun Oct 23 10:40:33 2016 -0700
>>> # Node ID 4c96a177e4ff6a63156ec8cf49483a78e4437756
>>> # Parent  0c41c0cb9b1ef7df40a30672927229ac195b1c92
>>> revlog: optimize _chunkraw when startrev==endrev
>>>
>>
>> LGTM, queued the series, thanks.
>>
>> In many cases, _chunkraw() is called with startrev==endrev. When
>>> this is true, we can avoid an extra index lookup and some other
>>> minor operations.
>>>
>>> On the mozilla-unified repo, `hg perfrevlogchunks -c` says this
>>> has the following impact:
>>>
>>> ! read w/ reused fd
>>> ! wall 0.371846 comb 0.370000 user 0.350000 sys 0.020000 (best of 27)
>>> ! wall 0.337930 comb 0.330000 user 0.300000 sys 0.030000 (best of 30)
>>>
>>> ! read batch w/ reused fd
>>> ! wall 0.014952 comb 0.020000 user 0.000000 sys 0.020000 (best of 197)
>>> ! wall 0.014866 comb 0.010000 user 0.000000 sys 0.010000 (best of 196)
>>>
>>> So, we've gone from ~25x slower than batch to ~22.5% slower.
>>>
>>
> I'm double reviewing this and I do not understand these number. I see
> where you 25x number come from, but I fail to get what this ~22.5% slower
> one is about. Can you clarify this ?
>

That "22.5%" is supposed to be "22.5x."


>
> At this point, there's probably not much else we can do except
>>> implement an optimized function in the index itself, including in C.
>>>
>>
> +1, if we get a significant high level performance boost it make sense to
> move more of the revlog code into C (or other compiled language).
>
> Cheers,
>
> --
> Pierre-Yves David
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.mercurial-scm.org/pipermail/mercurial-devel/attachments/20161106/a706bfc4/attachment.html>