[FFmpeg-devel] [PATCH 2/2] Add SIPR decoder for 5k0, 6k5 and 8k5 modes

Vitor Sessak vitor1001
Mon Jan 11 04:05:34 CET 2010


Vitor Sessak wrote:
> Vitor Sessak wrote:
>> Ronald S. Bultje wrote:
>>> Hi Vitor,
>>>
>>> On Jan 7, 2010, at 10:23 PM, Vitor Sessak <vitor1001 at gmail.com> wrote:
>>>> in:                      { 2, 2.2,  7,  9.5, 10  }
>>>> out:                     { 2,   4,  7,  9.5, 10  }
>>>> ff_acelp_reorder_lsf():  { 2,   4,  7,  9.5, 11.5}
>>>> sane:                    { 2,   4,  7,  9.5, 11.5}
>>>>
>>>> here we see a problem
>>>
>>> Ah, I see what you mean, min_dist is not enforced for the last lsf. 
>>> Very interesting. For WMAVoice, that's not the case, it's exactly 
>>> like the int version. OK, I guess that makes the discussion moot then.
>>>
>>> Does output quality audibly differ when adding in this "feature"? (It 
>>> might just be a decoder bug, WMAVoice is full of them.) Deviating 
>>> from binary can sometimes be OK...
>>>
>>> If the output doesn't change or worsens, patch OK (but please explain 
>>> this difference in a doxy); if it improves, it might be worth using 
>>> that instead.
>>
>> No, the output gets completely garbled if I do like in 
>> ff_acelp_reorder_lsf().
>>
>> Does anyone else wants to comment/review this codec? If nobody comment 
>> in a couple of days I'll commit it.
> 
> Committed.
> 
> Now the patch that adds 16k support. I'm undecided if I should put this 
> code in a separated file...

I've just noticed that this decoder depends on a patch that I forgot to 
attach. It that add floating point versions of some fixed point code 
already in SVN.

-Vitor
-------------- next part --------------
A non-text attachment was scrubbed...
Name: float_g729_code.diff
Type: text/x-patch
Size: 3176 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100110/d0da1691/attachment.bin>



More information about the ffmpeg-devel mailing list