On Wed, 7 Jun 2017 16:34:46 +0200
Steve Lhomme <[email protected]> wrote:

> On Wed, Jun 7, 2017 at 12:13 PM, Hendrik Leppkes <[email protected]> wrote:
> > On Wed, Jun 7, 2017 at 9:10 AM, Steve Lhomme <[email protected]> wrote:  
> >> On Tue, Jun 6, 2017 at 6:51 PM, wm4 <[email protected]> wrote:  
> >>> This also adds support to avconv (which is trivial due to the new
> >>> hwaccel API being generic enough). For now, this keeps avconv_dxva2.c as
> >>> "dxva2-old", although it doesn't work as avconv.c can't handle multiple
> >>> hwaccels with the same pixfmt. It will be removed in a later commit.
> >>>
> >>> The new decoder setup code in dxva2.c is significantly based on work by
> >>> Steve Lhomme <[email protected]>, but with heavy changes/rewrites.
> >>> ---
> >>> diff --git a/libavcodec/hevcdec.c b/libavcodec/hevcdec.c
> >>> index 69d5908551..7a9182af9b 100644
> >>> --- a/libavcodec/hevcdec.c
> >>> +++ b/libavcodec/hevcdec.c
> >>> @@ -383,7 +383,7 @@ static void export_stream_params(AVCodecContext 
> >>> *avctx, const HEVCParamSets *ps,
> >>>
> >>>  static enum AVPixelFormat get_format(HEVCContext *s, const HEVCSPS *sps)
> >>>  {
> >>> -    #define HWACCEL_MAX (CONFIG_HEVC_DXVA2_HWACCEL + 
> >>> CONFIG_HEVC_D3D11VA_HWACCEL + \
> >>> +    #define HWACCEL_MAX (CONFIG_HEVC_DXVA2_HWACCEL + 
> >>> CONFIG_HEVC_D3D11VA_HWACCEL * 2 + \
> >>>                           CONFIG_HEVC_VAAPI_HWACCEL + 
> >>> CONFIG_HEVC_VDPAU_HWACCEL)
> >>>      enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmt = pix_fmts;
> >>>
> >>> @@ -391,6 +391,7 @@ static enum AVPixelFormat get_format(HEVCContext *s, 
> >>> const HEVCSPS *sps)
> >>>          sps->pix_fmt == AV_PIX_FMT_YUV420P10) {
> >>>  #if CONFIG_HEVC_D3D11VA_HWACCEL
> >>>          *fmt++ = AV_PIX_FMT_D3D11VA_VLD;
> >>> +        *fmt++ = AV_PIX_FMT_D3D11;  
> >>
> >> Can this be configured when building ? Since I will not support the
> >> new format anytime soon I'd rather not try this for every file we
> >> decode for nothing.
> >>  
> >
> > The old version is first in the format list, so your app gets called
> > with that first. Additional ifdefs everywhere and a configure option
> > would get seriously messy, just to avoid one if check in the app that
> > you should be having anyway (and possibly even do already?). Please
> > don't :)  
> 
> Yes but in VLC if you select no HW acceleration we still try all
> output formats until we find a matching configuration.

I mean, it would be simple to just check whether the AV_PIX_FMT maps to
something potentially useful in VLC, in the get_format callback? Even
if it's just a "if (fmt != AV_PIX_FMT_D3D11)".

> Also if the new format is better, shouldn't it come before the old one
> ? (I wouldn't mind having D3D11 above DXVA2 as well)

Hm, maybe. I feel like it shouldn't depend on the order, but at least
there's nothing that would speed against it.

So:

AV_PIX_FMT_D3D11
AV_PIX_FMT_D3D11VA_VLD
AV_PIX_FMT_DXVA2_VLD
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to