EGL_MESA_screen_surface proposal

[Adam Jackson]

On Wednesday 16 March 2005 14:46, Jon Smirl wrote:
> On Wed, 16 Mar 2005 13:15:23 -0500, Jon Smirl <jonsmirl at gmail.com> wrote:
> > The root of the problem is that there are more attributes on the mode
> > than are exposed in EGL. You could do a elgChooseMode(1024,768) and
> > get back 10 identical modes at the EGL level, but they would all be
> > different at the hardware level. EGL will pick the what it thinks is
> > the best mode and you won't be able to get to the others since you
> > can't tell them appart.
>
> I think the key problem here is that elgChooseMode(1024,768) can
> return more than one mode. I don't think you can make the argument
> that you need to return more than one mode at each resolution. You
> will want this to be the best mode possible, why would you want to set
> an inferior one? But not enough information is returned through the
> EGL API to actually determine what the best mode is at the given
> resolution.

What did I just say?  If you have omitted an attribute from the mode 
specification then you have implicitly stated that that's an axis where you 
_don't_ _care_.  And all the modes returned are equally "best" based on the 
selection criteria you provided.

> My first approach was to expand the amount of information available
> about the modes. Now instead I would propose that the internal
> implementation of the egl API should sort through the modes in all of
> their gory detail and pick the best one. Then only this one would be
> exposed via the egl API to upper layers.

Brian's API allows for both:

- Automatic selection based on a list of criteria and a default sort order
- Manual selection based on application logic

What exactly are you arguing _for_?  And how does the API Brian proposed not 
provide it?

> Doesn't this simplify everything?

No.  It simplifies selection for the dumb end user.  It complicates the driver 
level and the specification, and hides valid options.

If you want a smarter mode selection heuristic, put it in the server.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.freedesktop.org/archives/dri-egl/attachments/20050316/c06a42fd/attachment.pgp