[FFmpeg-devel] [PATCH 2/2] avcodev/libdcadec: implement request_channel_layout

[James Almer]

On 09/07/15 9:10 AM, Hendrik Leppkes wrote:
> On Thu, Jul 9, 2015 at 1:32 PM, wm4 <nfxjfg at googlemail.com> wrote:
>> On Tue,  7 Jul 2015 03:06:24 -0300
>> James Almer <jamrial at gmail.com> wrote:
>>
>>> Signed-off-by: James Almer <jamrial at gmail.com>
>>> ---
>>>  libavcodec/libdcadec.c | 34 ++++++++++++++++++++++++++++++++++
>>>  1 file changed, 34 insertions(+)
>>>
>>> diff --git a/libavcodec/libdcadec.c b/libavcodec/libdcadec.c
>>> index 4bc3765..796826e 100644
>>> --- a/libavcodec/libdcadec.c
>>> +++ b/libavcodec/libdcadec.c
>>> @@ -36,6 +36,8 @@ typedef struct DCADecContext {
>>>      int buffer_size;
>>>  } DCADecContext;
>>>
>>> +static int downmix_warned = 0;
>>> +
>>>  static int dcadec_decode_frame(AVCodecContext *avctx, void *data,
>>>                                 int *got_frame_ptr, AVPacket *avpkt)
>>>  {
>>> @@ -131,6 +133,17 @@ static int dcadec_decode_frame(AVCodecContext *avctx, void *data,
>>>      if (exss = dcadec_context_get_exss_info(s->ctx)) {
>>>          enum AVMatrixEncoding matrix_encoding = AV_MATRIX_ENCODING_NONE;
>>>
>>> +        if (!downmix_warned) {
>>> +            uint64_t layout = avctx->request_channel_layout;
>>> +
>>> +            if (((layout == AV_CH_LAYOUT_STEREO_DOWNMIX || layout == AV_CH_LAYOUT_STEREO) && !exss->embedded_stereo) ||
>>> +                ( layout == AV_CH_LAYOUT_5POINT1 && !exss->embedded_6ch))
>>> +                av_log(avctx, AV_LOG_WARNING, "%s downmix was requested but no custom coefficients are available, "
>>> +                                              "this may result in clipping\n",
>>> +                                              layout == AV_CH_LAYOUT_5POINT1 ? "5.1" : "stereo");
>>> +            downmix_warned = 1;
>>> +        }
>>> +
>>
>> What does this mean? Does this happen with FFmpeg's DCA decoder too?
>>
> 
> It just means that the DCA stream has no embedded downmix
> coefficients. It'll downmix using the default coeffs, which can mean
> two things, 1) this was intended, and all is fine, 2) downmixing was
> not considered, and it may clip.
> Personally, I might like a mode where it would downmix if the coeffs
> are available, and not downmix if not (so I can downmix clipping-safe
> using av/swresample later).

It can't be currently done, as the behavior of the flags is pretty inconsistent
as i described in the patch.

We could however open a ticket asking the dcadec developer to actually change
the implementation so something like that is possible before the first version
is tagged (To avoid ifdeffery).

> 
> The native decoder would have the same problem, but I don't know if it
> even supports embedded downmixes properly.
> 
> - Hendrik
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>