[xine-devel] Re: VIDIX driver for xine

[Nick K. <ni...@ra...>]

Hello, Miguel!

On 30 May 2002 21:23:03 -0300 you wrote:

> On Tue, 2002-05-28 at 17:24, Gabucino wrote: 
> > And xine just plain sux for me. No TVout, mga_vid, I don't need GUI, buggy,
> > etc..
> ... one hacking-hour latter ... ;-) 
> 
> hi folks, 
> 
> I finished a very first version of a vidix driver for xine. For those
> who don't know what i'm talking about, vidix is a library by Nick
> Kurshev enabling direct access to video cards memory and back end
> scaler. I guess it also happens to be the only (?) modular piece of code
> from mplayer... 
> 
> Actually i'm not sure if we should include it in our cvs, so i'd like to
> hear opinions. 
But in this case xine users will need download and install mplayerxp first :)
> 
> Pros of video_out_vidix: 
> - faster video blitting compared to Xv (sometimes very noticeable
> difference) 
> - immediate support of several hardware (mga_vid, nvidia, pm3, radeon
> and mach64) 
> - no specific hardware code in xine (the api is simple and provides a
> nice abstraction layer) 
> 
> Cons of video_out_vidix: 
> - requires root priviledges to run (maybe there is some way to avoid
> this ?) 
> - ?? 
Of course - there was implemented dhahelper to avoid this problem
(control port and memory access through user/group permissions)
but it's still in alpha stage and is not portable.
As for me - I want to improve that in way of busmatsering support only.
(under linux, of course)
> 
> To test this driver you will need to install vidix in your computer.
> normal mplayer's "make install" should do the job installing the drivers
> but it doesn't seem to copy libvidix.a. In fact i'm testing it using a
> shared library libvidix.so. 
> 
Why do you need libvidix.so?
IMHO, this stuff is too small to be implemented as shared object.
Although it's possible too :)
> (btw, i don't know if vidix is actually hosted on mplayer project or
> mplayerxp) 
> 
It seems that on mplayerxp ;)
> Some todo list if we decide to include this driver: 
> - finish detection of vidix capabilities, including upscale and
> downscale flags and supported frame formats. 
> - implement a configure detection to compile it only when available (we
> might probably ask Nick to install libvidix and headers in some standard
> place to use) 
> 
which standard place do you want?
and where should be located its drivers in this case?
> 
> regards, 
> 
> Miguel 
> 
> 


Best regards! Nick