[x265] fix weightCost() [CHANGES OUTPUT]

[Steve Borho]

On 01/06, Satoshi Nakagawa wrote:
> # HG changeset patch
> # User Satoshi Nakagawa <nakagawa424 at oki.com>
> # Date 1420511389 -32400
> #      Tue Jan 06 11:29:49 2015 +0900
> # Node ID a260403b0d21cd2948fb2546997269c102249369
> # Parent  f255e8d06423231cb8c58ab5d3b10de7fb27b424
> fix weightCost() [CHANGES OUTPUT]
> 
> diff -r f255e8d06423 -r a260403b0d21 source/encoder/weightPrediction.cpp
> --- a/source/encoder/weightPrediction.cpp	Fri Jan 02 18:22:38 2015 +0530
> +++ b/source/encoder/weightPrediction.cpp	Tue Jan 06 11:29:49 2015 +0900
> @@ -193,9 +193,9 @@
>      if (bLuma)
>      {
>          int cu = 0;
> -        for (int y = 8; y < height; y += 8, r += 8 * stride, f += 8 * stride)
> +        for (int y = 0; y < height; y += 8, r += 8 * stride, f += 8 * stride)

I thought maybe x264 was skipping the top/bottom rows here like they do
for estimated frame cost, but it turns out they don't.  I'm not sure why
we ever did.

queued

>          {
> -            for (int x = 8; x < width; x += 8, cu++)
> +            for (int x = 0; x < width; x += 8, cu++)
>              {
>                  int cmp = primitives.satd[LUMA_8x8](r + x, stride, f + x, stride);
>                  cost += X265_MIN(cmp, cache.intraCost[cu]);
> @@ -203,12 +203,12 @@
>          }
>      }
>      else if (cache.csp == X265_CSP_I444)
> -        for (int y = 16; y < height; y += 16, r += 16 * stride, f += 16 * stride)
> -            for (int x = 16; x < width; x += 16)
> +        for (int y = 0; y < height; y += 16, r += 16 * stride, f += 16 * stride)
> +            for (int x = 0; x < width; x += 16)
>                  cost += primitives.satd[LUMA_16x16](r + x, stride, f + x, stride);
>      else
> -        for (int y = 8; y < height; y += 8, r += 8 * stride, f += 8 * stride)
> -            for (int x = 8; x < width; x += 8)
> +        for (int y = 0; y < height; y += 8, r += 8 * stride, f += 8 * stride)
> +            for (int x = 0; x < width; x += 8)
>                  cost += primitives.satd[LUMA_8x8](r + x, stride, f + x, stride);
>  
>      return cost;
> @@ -381,9 +381,9 @@
>                  break;
>  
>              case 2:
> -                fref = refFrame->m_fencPic->m_picOrg[2];
>                  orig = fencPic->m_picOrg[2];
>                  stride = fencPic->m_strideC;
> +                fref = refFrame->m_fencPic->m_picOrg[2];
>                  width =  ((fencPic->m_picWidth  >> 4) << 4) >> cache.hshift;
>                  height = ((fencPic->m_picHeight >> 4) << 4) >> cache.vshift;
>                  if (mvs)
> _______________________________________________
> x265-devel mailing list
> x265-devel at videolan.org
> https://mailman.videolan.org/listinfo/x265-devel

-- 
Steve Borho