[vlc-devel] Re: vlc: svn commit r12954 (md)

Jean-Paul Saman jean-paul.saman at planet.nl
Tue Oct 25 22:35:11 CEST 2005


Gildas Bazin wrote:

>On Tuesday 25 October 2005 12:06, Marian Durkovic wrote:
>  
>
>>>>modifications for RTP reordering:
>>>> * for RTP, prebuffering is done within RTP access
>>>> * reordering now working within this buffer
>>>> * solves synchro problems since packet rate is preserved 
>>>>        
>>>>
>>> I think the i_seqno should be removed from block_t (block_t is a 
>>>      
>>>
>generic
>  
>
>>>struct used everywhere).
>>> In the access it is perfectly valid to use i_dts or i_pts field for 
>>>yourself. (You can even use i_flags within the private part).
>>>      
>>>
>>I have defined it since there's nothing of uint16_t in block_t. There 
>>    
>>
>might 
>  
>
>>be several passes through the loops and I wanted to have the code as much 
>>optimised as possible. Is it that much offending?
>>
>>    
>>
>
>The reasoning is just that block_t is used everywhere in the code (input, 
>decoders, stream output, etc...). So by adding this field you might have 
>optimised the rtp access plugin but you have basically "unoptimized" 
>everything else.
>Ok, adding 1 field is not really a massive issue but still, I think we 
>should really keep block_t as simple and small as possible.
>
>The rtp access code is also far from being cpu intensive, so you shouldn't 
>really worry about that kind of optimisations there. They just won't make 
>any difference.
>
>--
>Gildas
>
I must agree with Gildas to keep block_t as simple as possible.

Grtz,
Jean-Paul Saman.

-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html



More information about the vlc-devel mailing list