[Ffmpeg-devel] [PATCH] lpcm 20 and 24 bit support in MPEG PS

[Baptiste Coudurier]

Michael Niedermayer wrote:
> On Tue, Jan 30, 2007 at 02:38:09PM +0100, Baptiste Coudurier wrote:
>> Hi
>> 
>> Michael Niedermayer wrote:
>>> Hi
>>> 
>>> On Fri, Jan 26, 2007 at 10:24:29AM -0000, M?ns Rullg?rd wrote:
>>>> Baptiste Coudurier said:
>>>>> Baptiste Coudurier wrote:
>>>>>> Baptiste Coudurier wrote:
>>>>>>> Baptiste Coudurier wrote:
>>>>>>>> Hi
>>>>>>>> 
>>>>>>>> M?ns Rullg?rd wrote:
>>>>>>>>> Baptiste Coudurier said:
>>>>>>>>>> Diego Biurrun wrote:
>>>>>>>>>>> Baptiste, patch forgotten?  :)
>>>>>>>>>>> 
>>>>>>>>>> Not really, Mans did not comment on it, and Michael
>>>>>>>>>> said that it should use AVBitStreamFilter. Atm
>>>>>>>>>> AVBitStreamFilter API has no support for demuxing,
>>>>>>>>>> nor demuxer activation cap, so ....
>>>>>>>>> I'm enjoying the California sun for a few more days.
>>>>>>>>> I'll look at and think about this and other things
>>>>>>>>> when I get back home.
>>>>>>>>> 
>>>>>>>> Any update about that patch ? It fixes a broken
>>>>>>>> behaviour and add feature. If it's not ok, I'll fix it
>>>>>>>> using another way.
>>>>>>>> 
>>>>>>> Ping. Ok solution is not beautiful, can we at least add
>>>>>>> the bits_per_sample check and error when not 16 bit ?
>>>>>>> 
>>>>>>> Applying and mentioning "XXX: use AVBistreamFilter" is a
>>>>>>> solution too and would add support for those streams.
>>>>>>> 
>>>>>> Ping....
>>>>>> 
>>>>> We did had a bug report today because of that issue.
>>>> I'm still waiting for Michael to implement or approve some
>>>> means of actually using the bitstream filter he says is the
>>>> proper way to handle this.
>>> i cant approve a way if noone suggests a way ... :)
>>> 
>>> and if you prefer this can also be implemented at the decoder
>>> side ... or you could even approve the patch which adds it into
>>> the demxuer but i dont think this is where the code should be
>>> 
>> if you want to wait for someone to implement AVBitstreamFilter
>> automatic insertion, that's fine by me.
> 
> i dont see why this should be delayed until automatic bitstream
> filter insertion is supported, rather the total opposite ... the more
> bitstream filters we have the more people will want them to be
> activated automatically so it would speed up the auto insert
> implementation :)
> 

actually there isn't any support at all for bitstream filters after
demuxing, and before decoding.

>> IMHO a decoder is not a good idea until good support for >16 bit
>> per sample audio is provided.
> 
> well and that will never be provided if there is no decoder which
> outputs 32bit or 24 bit ... its a chicken and egg problem one needs
> the other so why not write a decoder which outputs 32 or 24 bit?

First set avctx->sample_fmt in every decoder, then change output buffer
system because it can't reasonably be int16_t * anymore.
That should not require a decoder which outputs 32 or 24 bit per sample.

If someone wants to try...

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA
SMARTJOG S.A.                                    http://www.smartjog.com
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
Phone: +33 1 49966312