[FFmpeg-devel] [FFmpeg-devel-irc] IRC log for 2010-03-02

Måns Rullgård mans
Thu Mar 4 05:07:26 CET 2010


Alex Converse <alex.converse at gmail.com> writes:

> On Wed, Mar 3, 2010 at 10:20 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
>> On Wed, Mar 03, 2010 at 12:00:25AM +0000, irc at mansr.com wrote:
>> [...]
>>> [01:47:27] <DonDiego> peloverde: so what is missing for sbr to get merged?
>>> [01:48:48] <peloverde> I actually sent an e-mail out asking that, and got instructions on how to fix the demuxer issue
>>> [01:49:14] <peloverde> Basically I see two out stand complaints
>>> [01:49:29] <peloverde> 1) The demuxer/sample rate issue, certainly a valid concern
>>
>>> [01:49:48] <peloverde> 2) The filterbank could be more efficient
>>> [01:50:29] <peloverde> the filterbank is on the right order of complexity, I have no doubt that it could be made more efficient in C code and could also be SIMDed
>>
>> i do have some doubt that it cant just be done with half the operations
>> that said, thats not holding the code up from svn, because we dont know if
>> or how that could be done.
>> whats holding it up is the lack of any proper review, ill try to do that
>>
>> one thing that would help me is some text/spec that describes the algorithm
>> (that is sbr as a whole not the filterbank)
>>
>>
>>> [01:50:50] <peloverde> However i think requiring new SIMD code is really out of the scope of a new decoder
>>
>> i never asked for simd optims
>>
>>
>>> [01:51:07] <peloverde> and writing really fast scalar C is a waste of time
>>
>> didnt ask for that either
>
> I never said you did. In this case Mans was asking for SIMD.

You must have misunderstood.  I was asking for some loops to be broken
out so they *could* be simded easily.

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



More information about the ffmpeg-devel mailing list