On 30.03.2009 06:23, Dave Airlie wrote: > On Mon, Mar 30, 2009 at 12:42 PM, Dave Airlie <airl...@gmail.com> wrote: >> Does anyone remember why we've only done macro tiling on the radeon >> for the color buffer so far? >> > > /me discovers the reason ouch. > > So the 2D engine can't deal with a microtiled surface as a source, > > so short of using the 3D engine to do blits etc, microtiling isn't > probably going to be useful for color buffers. True. I remember though, I experimented with this for 3D, ignoring any 2D corruption (not to mention it didn't work for 3D neither cause r100/r200 can't scan out from microtiled buffers), and using microtiling for color buffers did not have any performance benefits quite the contrary it was definitely slower. It's possible though this has changed with r300 (though of course with things like render-to-texture it could potentially end up faster on those old chips as well in some situations).
Roland > > For texture tiling I'm going to have to with keeping two copies as > with this limitation we can't copy back from VRAM a micro-tiled > texture with the blitter. Again we'd have to use the 3D engine to do > it. You could untile manually though I agree that sucks. Using the 3D engine instead of blitter though probably wouldn't be unreasonable, I guess this kind of operation is limited by bandwidth anyway so performance should be the same. Roland ------------------------------------------------------------------------------ -- _______________________________________________ Dri-devel mailing list Dri-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dri-devel
