[FFmpeg-devel] [PATCH] configure option for 32-bit MPEG audio output

[Mike Melanson]

Michael Niedermayer wrote:
> On Wed, Feb 11, 2009 at 08:08:54PM +0100, Aurelien Jacobs wrote:
>> Michael Niedermayer wrote:
>>
>>> On Wed, Feb 11, 2009 at 11:46:07AM +0100, Diego Biurrun wrote:
>>>> On Tue, Feb 10, 2009 at 10:52:45PM -0800, Mike Melanson wrote:
>>>>> I want to start doing automated testing of MPEG-1 audio conformance  
>>>>> vectors via FATE using the officially sanctioned RMS method. To that  
>>>>> end, FFmpeg needs to have configure-time support for 32-bit audio  
>>>>> output. The code is already in FFmpeg but it requires setting  
>>>>> CONFIG_AUDIO_NONSHORT manually in libavcodec/mpegaudio.h. I want to  
>>>>> remedy that with this patch.
>>>> Looks fine, but..
>>>>
>>>>> --- libavcodec/mpegaudio.h	(revision 17146)
>>>>> +++ libavcodec/mpegaudio.h	(working copy)
>>>>> @@ -65,7 +63,7 @@
>>>>>  
>>>>> -#if CONFIG_MPEGAUDIO_HP && CONFIG_AUDIO_NONSHORT
>>>>> +#if CONFIG_MPEGAUDIO_HP && CONFIG_AUDIO_LONG
>>>> ..why not just
>>>>
>>>>   if CONFIG_AUDIO_LONG
>>>>
>>>> ?

That's just how it was when I go there. And CONFIG_MPEGAUDIO_HP is 
enabled by default anyway. Per the docs, disabling it sacrifices quality 
for better speed.

>>>> Otherwise you have to pass two configure options to enable this..
>>> is there any advantage of not enabling CONFIG_AUDIO_NONSHORT?
>> Last time I proposed to enable CONFIG_AUDIO_NONSHORT by default
>> the issue that was raised was the PSNR drop in regtest.
>> It is still here:
> 
> right i remember now :(

Meanwhile, I have a method (several, really) for testing MPEG-1 audio in 
FATE that does not rely on long audio. But it would be nice to have so 
that FATE can use the more accurate and sanctioned method.

-- 
     -Mike Melanson