Re: RandR 1.4 additional X server patches
On Sun, Dec 05, 2010 at 09:05:47PM -0800, Keith Packard wrote: I've added code to handle the separation screen and screen pixmap sizes, and fixed a couple of minor issues from the previous patches. [PATCH 1/3] DIX is responsible for ref counting scanout pixmaps. DDX was double ref-counting scanout pixmaps. DIX handles this already. [PATCH 2/3] Set sprite transforms from RRSetCrtcConfigs Oops -- I didn't manage to get the sprite transforms hooked up from the all-in-one configuration request. [PATCH 3/3] Separate out screen size and screen pixmap sizes in RRScreenSizeSet Here's the interesting bit -- this allows the screen size to be different from the screen pixmap size. It does this by placing the screen size in both the window dimensions as well as the borderSize and winSize regions within the root window. The screen pixmap dimensions are stored in the borderClip region, which is what is used to compute all of the nested window clipping data. This takes advantage of the existing code for forcing a complete recomputation of all screen clip regions on screen resize. It optimizes that path so that the regions are only recomputed when the *pixmap* changes size, and not when the screen changes size. This means that changing the screen size and leaving the screen pixmap alone will result in no repainting. It's probably the most fiddly of the whole patch sequence for PCP, but at least it's not huge. These changes, plus the ones you sent out earlier, plus I guess the two at git://people.freedesktop.org/~keithp/xserver master that I can't find emails for Reviewed-by: Aaron Plattner aplatt...@nvidia.com The change to decouple the screen and screen pixmap sizes scares me. We're going to have to go through our driver with a fine-toothed comb to make sure we use the right ones in all the right places. ___ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: RandR 1.4 additional X server patches
On Mon, 6 Dec 2010 14:36:43 -0800, Aaron Plattner aplatt...@nvidia.com wrote: The change to decouple the screen and screen pixmap sizes scares me. We're going to have to go through our driver with a fine-toothed comb to make sure we use the right ones in all the right places. Understood. You shouldn't be using borderSize and winSize for anything in your driver; borderClip is the clipping bounds of the root window. I'd be interested to know if you're using the others at all though. -- keith.pack...@intel.com pgppahnqzlA54.pgp Description: PGP signature ___ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
Re: RandR 1.4 additional X server patches
On Mon, Dec 06, 2010 at 02:44:08PM -0800, Keith Packard wrote: On Mon, 6 Dec 2010 14:36:43 -0800, Aaron Plattner aplatt...@nvidia.com wrote: The change to decouple the screen and screen pixmap sizes scares me. We're going to have to go through our driver with a fine-toothed comb to make sure we use the right ones in all the right places. Understood. You shouldn't be using borderSize and winSize for anything in your driver; borderClip is the clipping bounds of the root window. I'd be interested to know if you're using the others at all though. I meant the root window size vs. the screen pixmap size. The driver is a lot better about it than when I started at NVIDIA, so it may just work. :) I don't think we'll have any problems with the clip regions, that code is pretty self-contained. A quick grep shows that we don't use borderSize anywhere, but we do use winSize for some obscure overlay colormap code that's been around for ages. We use borderClip and clipList all over the place. ___ xorg-devel@lists.x.org: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel