Re: [Dri-devel] Levels of Support

[Kreuzritter2000 <kre...@gm...>]

> I think both Frank and Alan have valid points. It's nice to know more
> than works/doesn't, but all drivers will have problems. What I've seen
> on other projects, like emulators (WINE, Executor), is a way for users
> to report their own experiences. So people could enter:

> User:
> Application:
> Architecture:
> OS:
> Card:
> Version:
> Date:
> Score (1-10):
> Comments:

> The nice thing about this, is that you could then search on various
> parts you're interested in. What card plays tribes2 the best? How well
> the does V3 work on NetBSD on Intel? It could all be filled in by users
> and therefore be low maintenance. Although the score is somewhat
> subjective it isn't as open to interpretation as "Works/Doesn't." It
> also tracks over time, so you could see it evolve.

> Any other opinions/comments?

>  - |Daryll

It's a good idea.

But i want to add something.
It would be really cool if there would also be the possibly to insert a be=
nchmark
comparison between the Win95/98/ME drivers to the Linux DRI drivers for th=
e
same game.
Because that are the framerates we have to fight for.

It's ok when we know how good this card is in Tribes2 (for example) under
Linux, but it will help us more when we also know what the card is possibl=
e to do=20
when the drivers are really fast and tweaked, that's why we also need a Wi=
ndows95/98/Me comparison=20
(benchmark of that game).

So, this will also help developers which drivers needs more tuning work to=
 be done.
It doesn't help when the driver supports all possible features, it is also=
 necessary that
the driver can do all or many of those  features at a good  framerate spee=
d.


For example in the moment i could choose to play Unreal Tournament under l=
inux or under Windows with
my Voodoo3.

The video quality from UT is very good with the linux driver but the frame=
rates are only the half of
what the card is able to do under windows.
So for playing UT my game OS choice would be Windows at the moment.


I could accept framerates that are 10 % slower under linux than under wind=
ows to play it under linux.
But i don't like to play a game that is 20 % or more percent slower under =
Linux then under windows.

The result of this could be that we could have a kind of comparison table,=
 for example:

We set the value 100 % as the speed of the driver under Windows.=20
Then we would get the following video card table with the linux drivers (v=
alues are only examples, not true values)
=20
Radeon 102 %
I810 70 %
Voodoo3 62 %
G400 95 %

This would tell us for example that the Radeon driver would need no
tuning, because in this example it is faster then the windows driver.
Also we would see that the driver speed of the G400 would be acceptable,
so need no tuning work to to on it.
Then we see that the driver for the i810 could need a little optimization.
And we would see that the value of the Voodoo driver is really bad, so thi=
s
driver would need lot of workhours to get more out of it.

That's only a example, but we should know how fast/slow a driver
is compared of it's best possible value that is possible with the card wit=
h tunned
drivers.

Best Regards,
 Oliver C.




=20



=20




=20