[MPlayer-dev-eng] [RFD] Something...

[Daniel Egger]

Am Sam, 2003-04-12 um 22.43 schrieb D Richard Felker III:

> My point is not that you should be using a better vo. My point is that
> it's silly to compare performance on a bad-performance machine like
> yours where fast hardware with proper support in free software is not
> available.

You're simply wrong here. Hardware support is excellent, the ATI Radeon
Mobility is one of the best supported GPUs under Linux out there.
Also my PowerBook beats an Athlon XP-1700 in dnetc's RC5-72. This shows
that it's just a matter of writing good code and using all the
capabilities.

>  On my system, mplayer is probably 2-3x as fast as xine
> (however, I'm not going to install all the bloated dependencies and
> stuff just to test a bad movie player that's going to be too slow for
> me to use, so I don't have actual numbers).

Well, I would have benched it for you if the current mplayer wouldn't freak
out on my sample vob file with "Too many audio packets in the buffer".
The perceived performance is equal with both on an Athlon XP-1700. I
suppose the difference is more like a few percent in either direction
rather than the 200-300% you claim...

> I've never heard of a GPU fried by bad code. If it can be done, that's
> broken hardware design. Also if it can be done, why hasn't some doze
> virus author written a virus to fry your GPU already?? Smells like fud
> to me.

I've fried an ATI Rage 128. Ask James Simmons about his statistics, he
managed to kill a few more and different cards.

There are virii which can toast a machine, I'm not sure about GPUs right
now but it doesn't seem unlikely to me. The problem with virii is that 
there authors tend to implement damage routines which work on many
devices rather than relying on a very specific bug.

> Also with the incompetence of the xf86 developers, I would trust
> Nick's untested code more than their xv code!

Who the hell are you to call the developers of the most used application
on UNIX(-like) systems incompetent?

>  If there are problems in his code on ppc, they would probably just
> result in the hardware being entirely inaccessible; it's very doubtful
> that random bad things would happen.

Right. Sending arbitrary data over a PCI bus/AGP in priviledged mode
is absolutely riskless, especially when endianess issues have to be
taken care of and some transparent or not so transparent bridges will
behave "interestingly" in some cases....

Heck, even lspci -M or -xxx can crash a machine with some hardware and
they are well tested on a lot of systems.

-- 
Servus,
       Daniel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Dies ist ein digital signierter Nachrichtenteil
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/attachments/20030413/cdfbead3/attachment.pgp>