> On 13 Apr 2016, at 14:48, Mark Thompson <s...@jkqxz.net> wrote:
> 
> On 13/04/16 09:14, nablet developer wrote:
>> 
>>> On 13 Apr 2016, at 14:08, wm4 <nfx...@googlemail.com> wrote:
>>> 
>>> On Thu,  7 Apr 2016 11:44:20 -0400
>>> nablet developer <s...@nablet.com <mailto:s...@nablet.com>> wrote:
>>> 
>>>> Signed-off-by: nablet developer <s...@nablet.com>
>>>> ---
>>>> libavcodec/qsv.c          | 64 
>>>> +++++++++++++++++++++--------------------------
>>>> libavcodec/qsv.h          | 53 +++++++++++++++++++++++++++++++++++++++
>>>> libavcodec/qsv_api.c      | 26 +++++++++++++++++++
>>>> libavcodec/qsv_internal.h | 15 +----------
>>>> libavcodec/qsvdec.c       | 13 +++++-----
>>>> libavcodec/qsvdec.h       |  3 ++-
>>>> libavcodec/qsvenc.c       | 16 ++++++------
>>>> libavcodec/qsvenc.h       |  2 +-
>>>> 8 files changed, 125 insertions(+), 67 deletions(-)
>>> 
>>> Why would this API need to be exported?
>> 
>> previously QuickSync was used only by libavcodec and its components - e.g. 
>> there are QSV encoder and decoders for AVC and MPEG-2. so it was OK that 
>> QuickSync initialisation and cleanup functions were local for libavcodec.
>> 
>> but right now we're adding QuickSync VPP component to libavfilter, so 
>> mentioned functions now become shared at least between libavcodec and 
>> libavfilter.
>> therefore, patch to add QSV VPP filter was rejected because it accessed 
>> libavcodec functions which were local, and it was suggested that such 
>> functions are need to be exported from libavcodec, so libavfilter can use 
>> them.
>> 
> 
> This is precisely one of the problems that the hwcontext code was designed to
> solve.  I suggest using that rather than adding new ad-hoc codec-specific API
> calls - make libavutil/hwcontext_qsv.c; it should not require any new API 
> calls
> at all.  As a bonus, it also solves the context propagation problem which you
> will run into later when combining multiple filter and codec components.

okay, I will add libavutil/hwcontext_qsv and prepare new patch. thanks a lot 
for the suggestion.

> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Reply via email to