EGL_MESA_screen_surface proposal

Benjamin Herrenschmidt benh at kernel.crashing.org
Wed Mar 16 17:22:08 PST 2005


On Wed, 2005-03-16 at 13:26 -0500, Jon Smirl wrote:
> Another way to look at this is should EGL even expose more than one
> mode at any given resolution?
> If your card and monitor both support 1080p why would you ever want to
> use 1080i?
> If your card and monitor both support 1024x768-72 why would you want
> 1024x768-60?
> 
> Maybe I am going at this the wrong way, instead of making multiple
> modes at a resolution accessible maybe we should be hiding them.
> 
> Color depth is an independent variable to the mode.

Well, look like what Apple does in MacOS X, it's a good example of that.
They can have more than one mode in every resolution. They can mix in
the encoding scheme (pal/secam/ntsc) for TV modes, they can mix in the
way the stretcher is used on flat panels (keep aspect ratio or full
expansion, useful on non-4:3 flat panels), etc... They typically can
have up to 2 or 3 modes with the same size and different attributes that
will probably not be exposed by the EGL API (which is fine).

I concurr with Jon here that it would be useful to be able to obtain the
"user visible" name of a mode as a string, and the opposite function for
looking up a mode by it's user visible name. Software would be expected
to default to a size/freq lookup if there is no equivalent, it should be
made very precise in the API that no software is ever expected to "build
up" mode string names tho. The only way they might be manipulated is for
finding a mode from it's name when the name was itself obtained from EGL
earlier.

Ben.




More information about the dri-egl mailing list