coordinate transformation

[Timo Kreuzer]

Hi,

I'm currently working on the coordinate transformation code in win32k.
While Windows does this in win32k, I wonder why its not done in user 
mode and if we could depart from windows design here.

The advantages would be decreased complexity in the kernel and no use of 
floating point emulation on x86 (better performance).

Are there any things that would make it unreasonable?

I don't plan to go and change this now, I just like to discuss whether 
we might or might not do that at some point.

Timo

[James Tabor]

Hi!

On Wed, Jan 5, 2011 at 10:36 AM, Timo Kreuzer [e-mail] wrote:
> Hi,
>
> I'm currently working on the coordinate transformation code in win32k.
> While Windows does this in win32k, I wonder why its not done in user mode
> and if we could depart from windows design here.

Commercialization, the push to get a product out as fast as possible
and if it works don't break it. So, nothing changes.

You are seeing the same thing I did. If it can be done, yes!
> The advantages would be decreased complexity in the kernel and no use of
> floating point emulation on x86 (better performance).
>
> Are there any things that would make it unreasonable?

No, I'm doing it with user32 right now, settling issue where it can be
pushed to user or not.
> I don't plan to go and change this now, I just like to discuss whether we
> might or might not do that at some point.
>
> Timo

Example: with arwinss, why use wine handle allocation for gdi object
when ours works and replace the DCATTR with the wine structure and use
the DCATTR allocation scheme for the user wine structures. Same for
brush and the rest.

Everyone keeps forgetting the original mission here! If we can do it
better, then do it!
James