On Wed, 4 Oct 2017 23:04:04 +0100
Mark Thompson <s...@jkqxz.net> wrote:

> On 04/10/17 09:54, wm4 wrote:
> > On Wed,  4 Oct 2017 09:07:11 +0100
> > Mark Thompson <s...@jkqxz.net> wrote:
> >   
> >> ---
> >>  libavcodec/h263dec.c       | 10 ++++++++++
> >>  libavcodec/h264dec.c       | 19 +++++++++++++++++++
> >>  libavcodec/hevcdec.c       | 19 +++++++++++++++++++
> >>  libavcodec/hwaccel.h       | 27 +++++++++++++++++++++++++++
> >>  libavcodec/mmaldec.c       |  7 +++++++
> >>  libavcodec/mpeg12dec.c     | 24 +++++++++++++++++++++++-
> >>  libavcodec/mpeg4videodec.c | 10 ++++++++++
> >>  libavcodec/qsvdec.c        | 11 +++++++++++
> >>  libavcodec/qsvdec.h        |  2 ++
> >>  libavcodec/qsvdec_h2645.c  |  2 ++
> >>  libavcodec/qsvdec_other.c  |  3 +++
> >>  libavcodec/vc1dec.c        | 19 +++++++++++++++++++
> >>  libavcodec/vp8.c           |  7 +++++++
> >>  13 files changed, 159 insertions(+), 1 deletion(-)  
> > 
> > LGTM. The macros seem a bit messy on the first look, but I like that
> > the line noise/duplication is kept to a minimum.
> > 
> > I _think_ with this we could easily remove the get_format mess in the
> > decoders. See for example hevcdec.c/get_format(): we go through lots of
> > effort and ifdeffery to build the pix_fmts[] array, which seems dumb.
> > Just mentioning that in case someone complains that this patch adds
> > more ifdeffery.  
> 
> Not sure.  You have annoying cases with only subsets of formats (e.g. 8-bit 
> only or only YUV 4:2:0) being supported by the hardware APIs, which would 
> need to be expressed anyway.

The hwaccel implementation should check these things instead, at least
if there's in consistent support across underlying OS APIs.

> > Though I also wonder if we could merge the AVHWAccel entrypoints into
> > AVCodec and AVCodecHWConfig, which would require different definitions
> > again.  
> I have kindof done this part - in the next version, the only reference to an 
> AVHWAccel structure is in the corresponding CodecHWConfig.

Sounds good.
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to