[x265] [PATCH] Added support for multiple references

[Steve Borho]

On Wed, Sep 25, 2013 at 6:06 AM, Shazeb Khan <shazeb at multicorewareinc.com>wrote:

> # HG changeset patch
> # User Shazeb Nawaz Khan <shazeb at multicorewareinc.com>
> # Date 1380099889 -19800
> #      Wed Sep 25 14:34:49 2013 +0530
> # Node ID 0cc1b50c7d500c182332f09fc66420002909d5ec
> # Parent  4d0ced5c64fed290302b730ef8ac7d8613420a0d
> Added support for multiple references
>
> Added a commandline parameter '--ref <max number of references>'. DPB size
> adjusted to use number of references.
>

This patch was line wrap damaged, but I've queued a corrected version for
default.

As a follow-up, we need to display maxrefs in the logs, especially if it is
larger than 1.


>
> diff -r 4d0ced5c64fe -r 0cc1b50c7d50 source/common/common.cpp
> --- a/source/common/common.cpp Wed Sep 25 11:33:50 2013 +0530
> +++ b/source/common/common.cpp Wed Sep 25 14:34:49 2013 +0530
> @@ -160,7 +160,7 @@
>      param->bEnableSignHiding = 1;
>      param->bEnableTransformSkip = 1;
>      param->bEnableTSkipFast = 1;
> -
> +    param->maxNumReferences = 1;
>      /* Loop Filter */
>      param->bEnableLoopFilter = 1;
>
> diff -r 4d0ced5c64fe -r 0cc1b50c7d50 source/encoder/dpb.h
> --- a/source/encoder/dpb.h Wed Sep 25 11:33:50 2013 +0530
> +++ b/source/encoder/dpb.h Wed Sep 25 14:34:49 2013 +0530
> @@ -52,10 +52,9 @@
>          m_lastIDR = 0;
>          m_pocCRA = 0;
>          m_bRefreshPending = false;
> -        m_maxRefL0 = 1;             //TODO: This values should later be
> fetched from input params
> +        m_maxRefL0 = cfg->param.maxNumReferences;
>          m_maxRefL1 = 1;
>      }
> -
>      ~DPB();
>
>      void prepareEncode(TComPic*);
> diff -r 4d0ced5c64fe -r 0cc1b50c7d50 source/encoder/encoder.cpp
> --- a/source/encoder/encoder.cpp Wed Sep 25 11:33:50 2013 +0530
> +++ b/source/encoder/encoder.cpp Wed Sep 25 14:34:49 2013 +0530
> @@ -263,8 +263,8 @@
>      vps.setMaxLayers(1);
>      for (int i = 0; i < MAX_TLAYER; i++)
>      {
> -        m_numReorderPics[i] = _param->bframes;
> -        m_maxDecPicBuffering[i] = _param->bframes + 2;
> +        m_numReorderPics[i] = 1;
> +        m_maxDecPicBuffering[i] = X265_MIN(MAX_NUM_REF,
> X265_MAX(m_numReorderPics[i] + 1, _param->maxNumReferences));
>          vps.setNumReorderPics(m_numReorderPics[i], i);
>          vps.setMaxDecPicBuffering(m_maxDecPicBuffering[i], i);
>      }
> diff -r 4d0ced5c64fe -r 0cc1b50c7d50 source/x265.h
> --- a/source/x265.h Wed Sep 25 11:33:50 2013 +0530
> +++ b/source/x265.h Wed Sep 25 14:34:49 2013 +0530
> @@ -248,7 +248,7 @@
>      int       bEnableTransformSkip;            ///< enable intra
> transform skipping
>      int       bEnableTSkipFast;                ///< enable fast intra
> transform skipping
>      int       bEnableRDOQTS;                   ///< enable RD optimized
> quantization when transform skip is selected
> -
> +    int       maxNumReferences;                ///< maximum number of
> references a frame can have in L0
>      // loop filter
>      int       bEnableLoopFilter;               ///< enable Loop Filter
>
> diff -r 4d0ced5c64fe -r 0cc1b50c7d50 source/x265opts.h
> --- a/source/x265opts.h Wed Sep 25 11:33:50 2013 +0530
> +++ b/source/x265opts.h Wed Sep 25 14:34:49 2013 +0530
> @@ -66,6 +66,7 @@
>  OPT("bframes",         param->bframes,                required_argument,
> 'b', "Maximum number of consecutive b-frames (now it only enables B GOP
> structure)")
>  OPT("bframe-bias",     param->bFrameBias,             required_argument,
> 0, "Bias towards B frame decisions")
>  OPT("b-adapt",         param->bFrameAdaptive,         required_argument,
> 0, "0 - none, 1 - fast, 2 - full (trellis) adaptive B frame scheduling")
> +OPT("ref",             param->maxNumReferences,       required_argument,
> 0, "max number of references to be allowed (1 .. 16) ")
>  // Disabled because weighted uni-prediction was busted by not using
>  // pre-generated planes in motion compensation
>  //OPT("no-weightp",      param->bEnableWeightedPred,
>  no_argument, 0, "Disable weighted prediction in P slices")
>
>
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel
>
>


-- 
Steve Borho
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20130925/b5d6398c/attachment.html>