[x265] [PATCH] no-rdo: Replace optimized merge routine to decrease bitrate by 1.6%

[Deepthi Nandakumar]

This patch essentially reverts a merge optimization that was pushed a few
weeks ago.

The merge optimization included removing a cost estimate (merge mode with
residual) from consideration. This caused a bitrate/quality drop in rd 0/1
- this penalty needs to be investigated more before it becomes either a
user option or part of rd 0/1.


On Thu, Oct 3, 2013 at 11:22 AM, Steve Borho <steve at borho.org> wrote:

>
>
>
> On Thu, Oct 3, 2013 at 12:48 AM, <sumalatha at multicorewareinc.com> wrote:
>
>> # HG changeset patch
>> # User sumalatha polureddy
>> # Date 1380779314 -19800
>> # Node ID 942db71c75d2a6abb66bbca3d50fdb6a6a721b2c
>> # Parent  2c73823af5223e38d019dbbca56c2aa7fe58245e
>> no-rdo: Replace optimized merge routine to decrease bitrate by 1.6%
>>
>> In original merge routine, best candidate is selected based on the cost of
>> encoding with and without residue.
>> In optimized routine, the best merge candidate is selected based on cost
>> without residue which increased bitrate by 1.6%.
>> So replacing optimized routine with original routine to reduce the bitrate
>>
>
> What is the performance implication of this?  it sounds like a choice that
> should probably be left in the user's hands.
>
>
>> diff -r 2c73823af522 -r 942db71c75d2 source/encoder/compress.cpp
>> --- a/source/encoder/compress.cpp       Wed Oct 02 17:00:40 2013 +0530
>> +++ b/source/encoder/compress.cpp       Thu Oct 03 11:18:34 2013 +0530
>> @@ -412,7 +412,7 @@
>>
>>          /* Compute  Merge Cost */
>>          bool earlyDetectionSkip = false;
>> -        xComputeCostMerge2Nx2N(m_bestMergeCU[depth], m_mergeCU[depth],
>> &earlyDetectionSkip, m_modePredYuv[3][depth], m_bestMergeRecoYuv[depth]);
>> +        xCheckRDCostMerge2Nx2N(m_bestMergeCU[depth], m_mergeCU[depth],
>> &earlyDetectionSkip, m_modePredYuv[3][depth], m_bestMergeRecoYuv[depth]);
>>
>>          if (!earlyDetectionSkip)
>>          {
>> _______________________________________________
>> x265-devel mailing list
>> x265-devel at videolan.org
>> https://mailman.videolan.org/listinfo/x265-devel
>>
>
>
>
> --
> Steve Borho
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131003/7c47472c/attachment-0001.html>