Hmm, this might reveal a bug. addSplitCost is called for splitPred after the 4 sub-blocks have been evaluated. Why are we adding this cost again, especially for md.bestMode (which could be a non-split mode)?
On Fri, Apr 22, 2016 at 5:10 AM, <[email protected]> wrote: > # HG changeset patch > # User Sagar Kotecha<[email protected]> > # Date 1461327009 -19800 > # Fri Apr 22 17:40:09 2016 +0530 > # Node ID 3f2dfc9dad6e8756c4a9b5afd5b5f38516e04570 > # Parent 02d79be487d7f825c961d15535a8681a201da3b1 > analysis: do not add split cost if skip is chosen > > diff -r 02d79be487d7 -r 3f2dfc9dad6e source/encoder/analysis.cpp > --- a/source/encoder/analysis.cpp Sun Apr 17 21:07:28 2016 +0000 > +++ b/source/encoder/analysis.cpp Fri Apr 22 17:40:09 2016 +0530 > @@ -1329,7 +1329,7 @@ > if (m_bTryLossless) > tryLossless(cuGeom); > > - if (mightSplit) > + if (mightSplit && !foundSkip) > addSplitFlagCost(*md.bestMode, cuGeom.depth); > } > > @@ -1707,7 +1707,7 @@ > if (m_bTryLossless) > tryLossless(cuGeom); > > - if (mightSplit) > + if (mightSplit && !foundSkip) > addSplitFlagCost(*md.bestMode, cuGeom.depth); > } > > _______________________________________________ > x265-devel mailing list > [email protected] > https://mailman.videolan.org/listinfo/x265-devel > -- Deepthi Nandakumar Engineering Manager, x265 Multicoreware, Inc
_______________________________________________ x265-devel mailing list [email protected] https://mailman.videolan.org/listinfo/x265-devel
