[MPlayer-dev-eng] Re: [PATCH]H264 over rtsp
Carl Eugen Hoyos
cehoyos at ag.or.at
Mon Sep 11 00:41:40 CEST 2006
Hi!
Ross Finlayson <finlayson <at> live555.com> writes:
>
> Look, I don't care how many times you submit this patch, but my
> response remains the same each time.
To start: I know reasons why my patch must not necessarily be applied (and I
wrote them before), but I don't understand:
1) your reasons
2) why you never answered my emails with (obviously necessary) one-line
patches.
And in my last email, I wrote "for those who are interested" (or something
similar). I already knew you don't like the patch, but the source had changed
and this is an often requested feature, so I (thought I) had to update it. No
reason to flame me!
> For the third time:
>
> >Most of this looks OK, *except* for that hideous code that you've
> >added to the "getBuffer()" routine in "demux_rtp.cpp". I'm not sure
> >what that code is supposed to be doing, but it looks like it's
> >trying to do some filtering on the incoming data that should really
> >be done (if at all) by the H.264 decoding library instead. In any
> >case, the VLC code doesn't do anything like this, and neither should
> >MPlayer.
> >
>
> Note especially the last sentence.
Since you asked so often and so kindly, I spent the last night with vlc (and
unfortunately, that's less funny than it sounds).
Of course, it's not totally clear what you mean with "the VLC code doesn't do
anything like this", so I'll try to describe what it definitely does:
When it receives H264 packets over RTSP, vlc copies the whole frames twice(!)
with memcpy, not with two calls to memcpy, but over ten (!), depending on the
frames fragmentation. That's a few calls to memcpy for every frame in vlc's
code compared to one memcpy for the whole stream in my patch! While copying
(that's actually possible, since it's so many calls), vlc also parses the frames.
(That's where I stopped my research: I already know it works with much less
parsing.)
Do you really want me to implement the memcpy's and the parsing? After
following this list, I believe, especially the memcpy would find many friends
;-)
I will try to follow Reimars suggestion and post another version of the
(basically identical) patch. Please don't flame me anymore, I already
understand you won't apply it (just not exactly why).
If, on the other hand, you want me to send smaller patches, please review this
one:
http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/2006-August/045297.html
Thanks for your patience, regards, Carl Eugen
More information about the MPlayer-dev-eng
mailing list