Bind to FBO when using GPU composition
We'll manage the ANativeWindow directly to pull buffers for drawing.
Some additional minor changes have snuck in to clean up unnecessary GL
calls. Checking simpleperf there is now 0 GL overhead during HWC
composed frames.
Bug: 117680609
Change-Id: If639ce2c6cd600eabb957d278815f80b413d92b3
Test: libsurfaceflinger_unittest, SurfaceFlinger_test, go/wm-smoke
Test: Android Auto works in my car
Test: Manual inspection of simulated virtual displays
diff --git a/services/surfaceflinger/SurfaceFlinger.h b/services/surfaceflinger/SurfaceFlinger.h
index e449e20..ca40913 100644
--- a/services/surfaceflinger/SurfaceFlinger.h
+++ b/services/surfaceflinger/SurfaceFlinger.h
@@ -716,10 +716,10 @@
void doDebugFlashRegions(const sp<DisplayDevice>& display, bool repaintEverything);
void doTracing(const char* where);
void logLayerStats();
- void doDisplayComposition(const sp<const DisplayDevice>& display, const Region& dirtyRegion);
+ void doDisplayComposition(const sp<DisplayDevice>& display, const Region& dirtyRegion);
// This fails if using GL and the surface has been destroyed.
- bool doComposeSurfaces(const sp<const DisplayDevice>& display);
+ bool doComposeSurfaces(const sp<DisplayDevice>& display);
void postFramebuffer(const sp<DisplayDevice>& display);
void postFrame();