Re: Some perspective from the cheap seats...

DD> but its seems that the purpose of exposing more hardware
DD> functionality [in Render] got lost somewhere along the line.

David, could you please explain that?

As far as I can tell, Render is exporting some fairly primitive
graphic primitives that should be available in any resonable hardware.
It exports much more than that, sure, but there's nothing preventing
you from ignoring the extra functionality.

What I like in Render is stuff that could be readily be accelerated:

  - PushPixels, at very last exposed to the clients (PictOpOver, depth 1);
  - PushPixels through an alpha mask (PictOpOver, depth n);
  - Opaque polygons;
  - anti-aliased polygons (imprecise variant).

Granted, these operations are not accelerated right now, and some are
surprisingly slow in the current implementation (PictOpSrc, for
example, is orders of magnitude slower than CopyArea, to the extent of
being unusable, although there's no good reason it shouldn't dispatch
to the very same driver primitive).

I'll readily grant, also, that the current Render protocol provides no
hint to the client about which operations are likely to be usefully
fast -- in contrast to the hints provided by RandR (I'm not discussing
the criticisms I've heard about RandR, please let's not get sidetracked).

Notwithstanding the above, I just don't understand the claim that
Render is going against what the driver people want to see.

                                        Juliusz