[Ffmpeg-devel] [PATCH] RTCP valid receiver statistics....

rdm4 at martellventures.com rdm4
Fri Nov 3 05:12:09 CET 2006



On Tue Oct 31 15:59 , Michael Niedermayer <michaelni at gmx.at> sent:

>Hi
>
>On Tue, Oct 31, 2006 at 03:17:40PM -0600, Ryan Martell wrote:
>[...]
>> >[...]
>> >>+#if 0
>> >>+/**
>> >>+* This function is currently unused; without a valid local ntp  
>> >>time, I don't see how we could calculate the
>> >>+* difference between the arrival and sent timestamp.  As a  
>> >>result, the jitter and transit statistics values
>> >>+* never change.  I left this in in case someone else can see a  
>> >>way. (rdm)
>> >
>> >the system time should be used
>> 
>> really?  how do I get system time in 64 bits, without just taking  
>> time_t and shifting appropriately? 
>
>hmm i dont think there is a portable way (or at least iam not aware of one)
>on x86 rdtsc could be used, there are probably also a few gnu extensions with
>which you could get an accurate time though i dunno how accurate and all
>that wouldnz help non gnu non x86 ...
>iam not sure how important that jitter stuff is? getting an accurate 
>time will likely be more complicated then i thought at first, rdtsc
>would also require the ticks/sec to be known ...

I'll look into it; as long as it's invariant (as you noted below) it doesn't really matter how accurate it is, I guess.

>> I don't know what the server does  
>> with the jitter, but i would be afraid of screwing something up if  
>> the two machine's times aren't close.
>
>the jitter calculation in rfc1889:
>D(i,j)=(Rj-Ri)-(Sj-Si)=(Rj-Sj)-(Ri-Si)
>J=J+(|D(i-1,i)|-J)/16
>
>is invariant under adding a constant to either the source or destination
>time, so it doesnt matter if one is of by an hour or 5000 years
>but of course accuracy matters ...
>
>
>[...]
>> >[...]
>> >>+#if defined(DEBUG) && 0
>> >>-#if defined(DEBUG) || 1
>> >
>> >unrelated change?
>> 
>> Nope; I am replacing the old method of checking sequences with the  
>> new method.
>> 
>> This version of the patch strips it out. (and fixes the above, except  
>> system time, which I would like to put in a later patch to setup the  
>> jitter).
>
>ok and patch looks ok

Can someone apply it please?  My laptop was stolen, and so I have to rebuild from here!

Thanks!
-Ryan




More information about the ffmpeg-devel mailing list