[FFmpeg-devel] [PATCH] MP3 conformance in FATE

[Måns Rullgård]

Vitor Sessak <vitor1001 at gmail.com> writes:

> Vitor Sessak wrote:
>> On 08/03/2010 06:21 PM, Michael Niedermayer wrote:
>>> On Sat, Jul 31, 2010 at 12:50:27AM +0200, Vitor Sessak wrote:
>>>> $subj. I didn't succeeded in making all the tests pass. The output looks
>>>> like:
>>>>
>>>> compl.bit  stddev:    0.03 PSNR:125.77 MAXDIFF:    1 bytes:   497664/
>>>> 497664
>>>> he_32khz.bit  stddev:    0.05 PSNR:120.94 MAXDIFF:    1 bytes:   345600/
>>>> 343296
>>>> he_44khz.bit  stddev:    0.05 PSNR:120.93 MAXDIFF:    1 bytes:   944640/
>>>> 942336
>>>> he_48khz.bit  stddev:    0.05 PSNR:120.94 MAXDIFF:    1 bytes:   345600/
>>>> 343296
>>>> hecommon.bit  stddev:    0.04 PSNR:123.93 MAXDIFF:    1 bytes:   138240/
>>>> 133632
>>>> he_free.bit  stddev:    0.04 PSNR:123.93 MAXDIFF:    1 bytes:   138240/
>>>> 308736
>>
>>>> he_mode.bit  stddev: 3831.79 PSNR: 24.66 MAXDIFF:16941 bytes:   294892/
>>>> 523008
>> According to the spec:
>>
>>>    he_mode.bit     tests mono/stereo/MS-stereo/intensity-stereo
>> Note the size discrepancy.
>>
>>>> si.bit  stddev:    0.00 PSNR:140.64 MAXDIFF:    1 bytes:   271872/
>>>> 269568
>>>> si_block.bit  stddev:    0.00 PSNR:141.93 MAXDIFF:    1 bytes:   147456/
>>>> 145152
>>
>>>> si_huff.bit  stddev: 8733.73 PSNR: 17.51 MAXDIFF:65535 bytes:   172800/
>>>> 170496
>> This one:
>>
>>>    si_block.bit    tests the block-modes (long,start,short/mixed,stop)
>> and this
>>
>>>> sin1k0db.bit  stddev:17503.40 PSNR: 11.47 MAXDIFF:40705 bytes:  1456128/
>>>> 1451520
>> is supposed to test
>>
>>>    sin1k0db.bit    A 1kHz 0dB sinus. The Decoder should generate full
>>>                    scaled output with this bitstream
>> BTW, to make the fate tests cleaner, I'd suggest to apply the
>> attached patch to ffmpeg.c, to make the behavior of "-fs ss" to mean
>> "make output file of size equals or less than ss". In this case, to
>> generate the required 269568 samples for si.bit, one would pass "-fs
>> 269568" instead of "-fs 269560".
>
> Michael, ping this simple patch?
>
>> Also, new patch attached using new fate.mak schema.
>
> Mans?

You know how it works, and the fate bits look good.

-- 
M?ns Rullg?rd
mans at mansr.com