Add interface for doing asynchonous cursor layer updates to hwcomposer
Change-Id: If2ab523b400cfb586c8e1fed98f2ebe3538c0419
diff --git a/include/hardware/hwcomposer.h b/include/hardware/hwcomposer.h
index af04181..3dfb4fd 100644
--- a/include/hardware/hwcomposer.h
+++ b/include/hardware/hwcomposer.h
@@ -141,6 +141,13 @@
* as a solid color since the platform is not currently able to composite
* sideband layers with the GPU. This may be improved in future
* versions of the platform.
+ *
+ *
+ * HWC_CURSOR_OVERLAY
+ * Set by the HWC implementation during (*prepare)(), this value
+ * indicates the layer's composition will now be handled by the HWC.
+ * Additionally, the client can now asynchronously update the on-screen
+ * position of this layer using the setCursorPositionAsync() api.
*/
int32_t compositionType;
@@ -505,11 +512,12 @@
* (*prepare)() can be called more than once, the last call prevails.
*
* The HWC responds by setting the compositionType field in each layer to
- * either HWC_FRAMEBUFFER or HWC_OVERLAY. In the former case, the
- * composition for the layer is handled by SurfaceFlinger with OpenGL ES,
- * in the later case, the HWC will have to handle the layer's composition.
- * compositionType and hints are preserved between (*prepare)() calles
- * unless the HWC_GEOMETRY_CHANGED flag is set.
+ * either HWC_FRAMEBUFFER, HWC_OVERLAY, or HWC_CURSOR_OVERLAY. For the
+ * HWC_FRAMEBUFFER type, composition for the layer is handled by
+ * SurfaceFlinger with OpenGL ES. For the latter two overlay types,
+ * the HWC will have to handle the layer's composition. compositionType
+ * and hints are preserved between (*prepare)() calles unless the
+ * HWC_GEOMETRY_CHANGED flag is set.
*
* (*prepare)() is called with HWC_GEOMETRY_CHANGED to indicate that the
* list's geometry has changed, that is, when more than just the buffer's
@@ -751,11 +759,37 @@
*/
int (*setActiveConfig)(struct hwc_composer_device_1* dev, int disp,
int index);
+ /*
+ * Asynchronously update the location of the cursor layer.
+ *
+ * Within the standard prepare()/set() composition loop, the client
+ * (surfaceflinger) can request that a given layer uses dedicated cursor
+ * composition hardware by specifiying the HWC_IS_CURSOR_LAYER flag. Only
+ * one layer per display can have this flag set. If the layer is suitable
+ * for the platform's cursor hardware, hwcomposer will return from prepare()
+ * a composition type of HWC_CURSOR_OVERLAY for that layer. This indicates
+ * not only that the client is not responsible for compositing that layer,
+ * but also that the client can continue to update the position of that layer
+ * after a call to set(). This can reduce the visible latency of mouse
+ * movement to visible, on-screen cursor updates. Calls to
+ * setCursorPositionAsync() may be made from a different thread doing the
+ * prepare()/set() composition loop, but care must be taken to not interleave
+ * calls of setCursorPositionAsync() between calls of set()/prepare().
+ *
+ * Notes:
+ * - Only one layer per display can be specified as a cursor layer with
+ * HWC_IS_CURSOR_LAYER.
+ * - hwcomposer will only return one layer per display as HWC_CURSOR_OVERLAY
+ * - This returns 0 on success or -errno on error.
+ * - This field is optional for HWC_DEVICE_API_VERSION_1_4 and later. It
+ * should be null for previous versions.
+ */
+ int (*setCursorPositionAsync)(struct hwc_composer_device_1 *dev, int disp, int x_pos, int y_pos);
/*
* Reserved for future use. Must be NULL.
*/
- void* reserved_proc[2];
+ void* reserved_proc[1];
} hwc_composer_device_1_t;