Test: Sample output is below
D/OpenGLRenderer: Resource Cache Usage:
D/OpenGLRenderer: 32 items out of 8192 maximum items
D/OpenGLRenderer: 4635572 bytes (4.42 MB) out of 96.00 MB maximum
This is less verbose than OpenGL memory debug output for
two reasons:
(1) SkiaGL has less caches.
(2) SkiaGL does not support printing on cache additions/evictions.
This seems like more of an internal debugging tool rather than
a user-facing debug feature. I think it's best to leave this
unimplemented until we find that it might be useful.
BUG:32370375
Change-Id: Ib063f1c2a7f88e9840341b1001d227f556d88f26
This adds support for debug.hwui.profile and
debug.hwui.show_dirty_regions to the Skia pipelines.
There still may be some follow up work for profiling
with visual bars. The speed at which the renderer
is able to draw the rects in the graph is having a
noticeable effect on the actual data in the graphs.
Test: Verified that the features work as expected.
BUG:32370375
Change-Id: I24430da2bab3eb54f6a771f9c984f8ae0008a5a7
Fix RenderNodeDrawable to draw nonzero Z nodes if not in a reordering
section.
Write an unit test modeled after FrameBuilder zReorder, which
verifies the bug fix.
Test: built and run unit tests on angler-eng.
bug: 32541103
Change-Id: Ifbf2d51f4432f5de3af4abe5987c2a72fed14185
Implement Skia pipelines for OpenGL and Vulkan:
base SkiaPipeline, SkiaOpenGLPipeline and SkiaVulkanPipeline.
Write unit tests for SkiaPipeline.
Test: Built and run manually on angler-eng.
Change-Id: Ie02583426cb3547541ad9bf91700602a6163ff58
Implement SkiaRecordingCanvas, RenderNodeDrawable, GLFunctorDrawable,
LayerDrawable, StartReorderBarrierDrawable, EndReorderBarrierDrawable.
Move AnimatedRoundRect and AnimatedCircle in a separate file.
All Skia pipeline files are moved in hwui/pipeline/skia folder.
Add unit tests for RenderNodeDrawable, StartReorderBarrierDrawable,
EndReorderBarrierDrawable and SkiaRecordingCanvas.
Test: I tested manually on 6P devices and did run the unit tests.
Change-Id: If2a347bd1fc4689953822294ce5bf98c7f3f57c7
