summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJakob Bornecrantz <wallbraker@gmail.com>2010-02-15 02:09:35 +0000
committerJakob Bornecrantz <wallbraker@gmail.com>2010-02-15 02:09:35 +0000
commit453dbedd61d143dd81b02479fb3e9778e1bace4d (patch)
tree1bd84a87d8b2b2aee73ddaf9e2b97baacf0836c4
parent6ad834b39d6c2ae9ead2e2b00908ad2fa6914897 (diff)
downloadmesa-gallium-sw-api.tar.gz
gallium: More comments about the drm api vs software rasterizersgallium-sw-api
-rw-r--r--src/gallium/include/state_tracker/xm_winsys.h24
1 files changed, 18 insertions, 6 deletions
diff --git a/src/gallium/include/state_tracker/xm_winsys.h b/src/gallium/include/state_tracker/xm_winsys.h
index 5e2b5ee8be1..07e0447cd4b 100644
--- a/src/gallium/include/state_tracker/xm_winsys.h
+++ b/src/gallium/include/state_tracker/xm_winsys.h
@@ -34,19 +34,31 @@ struct pipe_screen;
struct pipe_surface;
-/* One unusual thing about the way the dri2 driver api works is that
- * textures for scanout are not created through the normal
- * pipe_screen::create_texture() method, but rather by converting
- * existing DRM buffer handles into pipe_textures by a backdoor in the
- * driver, exposed through the winsys and drm_api callbacks.
+/* The drm driver stack have two different users st/dri and st/xorg.
*
+ * st/xorg produces both scanouts (primary) and shared (displaytarget)
+ * textures via the normal pipe_screen::create_texture() callback,
+ * this follows how the lp_winsys works. However the lp_winsys
+ * does not deal with the difference between scanouts and displaytargets
+ * something that needs to be added. st/xorg then uses the drm_api to
+ * access the handles to textures.
+ *
+ * st/dri creates texture via the drm api.
+ *
+ *
+ * With all of the above requirements the drm API needs to be aware that
+ * a software rasterizer has been layered ontop of it. However this code
+ * can be completely generic and reused for all drm winsys.
+ */
+
+/*
* In the software drivers, we would also like the co-state tracker to
* be involved in creating the backing for scanout textures. This
* allows the knowledge of eg. XShm to be collapsed down to a single
* location, and permits a null or at least tiny software rasterizer
* winsys.
*
- * The main question is whether to follow the approach of the DRI2
+ * The main question is whether to follow the approach of the st/dri
* state tracker and ask the driver to turn some pre-existing storage
* into a texture, or to stay closer to the lp_winsys approach and
* have the provide the driver with a set of callbacks allowing it to