Add ATRACE memory counters to track memory usage at the end
of each frame in HWUI.
There are 3 catagories: CPU, GPU and Texture memory.
There are 3 more counters for memory that can be purged.
This CL deletes GpuMemoryTracker class, which implemented
similar function for Android O HWUI renderer.
Test: Collected systrace with gmail and setting.
Test: memory tracing adds ~0.1ms per frame when ATRACE is enabled
Bug: 146580770
Change-Id: Icbcc0478bc426dff578e83726fe7c95df171ed93
By avoiding setting a dequeue buffer timeout we avoid
hitting a different path in BufferQueue that prevents
async behavior from happening. This restores P's
behavior in this path.
Bug: 143860379
Test: repro app in bug
Change-Id: Iffbd9f9e6689a40876ff3aa74c10020e3f09fc6a
Invoke Surface::enableFrameTimestamps after eglDestroySurface.
eglDestroySurface internally disables time stats. Order is
important, when CanvasContext::setSurface is invoked with
a surface, that is already current (which happens all the time).
Test: ran UiBench microbenchmark tests
Change-Id: I3d023c3a87da6329c556426d553c744e541b9dff
A new instance snuck in since I haven't completed the platform toolchain
upgrade yet. There was imprecision in this call; accept it by making the
implicit cast explicit.
Bug: 139945549
Test: mm
Change-Id: Id704d9741b480eba4cf3955e70119d935776faf6
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
We only need to redraw if the size /changed/ not
if it was the same. Also fix damageId to not use
frameNumber as repeated redraws of the same frame
would toggle.
Bug: 143711430
Test: systrace
Change-Id: I8ac4629c9ff4fd51de33d1be7aa46ccc995ba342
gui/Surface will eventually be removed since HWUI needs to depend on a
stable ABI, so use the NDK methods for querying width and height that
operates on ANativeWindow instead.
Bug: 137012798
Test: builds
Change-Id: I1309e31992190e7b44c6ea83f962f372b6b0afcf
This also removes the dependency on ui/DeviceInfo other than in test
code.
Bug: 136263392
Bug: 136263238
Test: builds, boots
Change-Id: I6a4687e981359f0e6beb83be8a5501ed7fd16f15
Poking around in a few apps it doesn't appear that
the VectorDrawableAtlas is achieving sufficient
utilization to justify its existence. The potential for
draw call merging doesn't seem warranted for the
RAM cost of the atlas.
Bug: 137853925
Test: builds
Change-Id: Id2419bc6dccb6316636d50c568f8fac75a2d563f
Now that they return nanoseconds, there's no reason to lose precision.
Bug: 137012798
Test: builds
Change-Id: I52ca20cb010b4f1829596e0a765b7013ae3665ca
Replaces the query() call for getting the duration of the internal
dequeue() call a call that will be abi stable.
Bug: 137012798
Test: builds
Change-Id: I6d2118a7c093dd2ecc4ebfb4b4ee8556f7e709a7
Handle TIMED_OUT better by rescheduling (TODO: give up after N
attempts?)
Fix SYNC_SURFACE_LOST_REWARD_IF_FOUND path to actually go fetch
a new surface.
Bug: 137509524
Test: Injected errors randomly, verified nothing got permanently dead.
Change-Id: Id30f8ad1dd7196041ee84c16c8cf5c814002a6ce
Timeout recovery isn't working properly, so increase it
to avoid any false-positives from something like a
slow screen-on
Bug: 137509524
Test: none
Change-Id: Ic4fae04b41f71600479f9557e334f67c27851db4
Using SYSTEM_TIME_MONOTONIC works for Android (where it translates to
CLOCK_MONOTONIC) and host targets, while CLOCK_MONOTONIC is not defined
on macOS.
Bug: 117921091
Test: existing tests should pass
Change-Id: I1fad472881830fb0701a320cf37319e083932ad4
This reverts commit f802abe47f.
Reason for revert: We're seeing jank increase after this patch.
Change-Id: Ic8d77df3fbbeeedeed4b8f31737454d9c056f8b0
Bug: 132733305
ro.hwui.render_ahead was behaving the same as -1 when it was set
to zero. Correct it so that ro.hwui.render_ahead will queue frames
when set to zero on any display rate FPS.
Fixes: 131340980
Test: Test presence or not of timestamps with -1,0,1,2 settings
Change-Id: Ibd1974253e947130fb03a0b86f599e0e5f5eac38
Tracks refresh rate changes and adjusts renderahead
based off of the active refresh rate.
Default is 60hz = 0 render ahead & > 70hz is render ahead 1
Bug: 127822449
Test: systraced stuff
Change-Id: I9849aa065262f21f7602d44cd1761373279dc28d
Bug: 128805564
Test: Manual + systrace; hwui_unit_tests; CtsUiRenderingTestCases
Only update a VectorDrawable's cache if it is onscreen. This fixes a
Twitter use case where the app has a ProgressBar that is exactly one
pixel offscreen. Prior to this CL, we repeatedly drew the ProgressBar's
AVD to a GPU surface, even though we clip it out later and never draw
that GPU surface. Now, we recognize that the AVD is outside of the
bounds of the screen, so we never draw to the GPU surface.
TreeInfo:
- store the size of the screen, retrieved from
CanvasContext::getNextFrameSize.
SkiaDisplayList:
- Store the matrix at the time of recording a VectorDrawable. Concat
that with the current matrix to determine whether the VD is on screen,
based on the TreeInfo. If it is offscreen, do not add it to the list
of AVDs that need to be updated ahead of rendering.
- In addition, if it is offscreen (or not dirty), do not call
setPropertyChangeWillBeConsumed(true). This prevents triggering
dispatchFrameCallbacks to update on the RenderThread when there is no
need to. This also mimics what would happen if the View/RenderNode had
been completely offscreen.
- Add a method to append an AVD to mVectorDrawables. Now that the vector
is of Pairs, this simplifies the call sites. Add a second helper to
just add an AVD without a matrix, for use in tests.
SkiaRecordingCanvas:
- get the current matrix and store it in the display list along with the
AVD.
CanvasContext:
- add getNextFrameSize, for reporting the size of the next frame without
dequeuing it
VectorDrawable.cpp:
- call quickReject to potentially short circuit drawing. This is for a
hypothetical use case (verified in a test app) where the containing
RenderNode is partially onscreen, but the AVD itself is not. Even
without the change to VectorDrawable.cpp, we skip uploading to the GPU
cache, the SkiaDisplayList still attempts to draw it. This change
keeps us from drawing it at all.
SkiaDisplayListTests.cpp:
- Now that I've hidden mVectorDrawables, call the new public APIs.
- prepareListAndChildren test
- for the clean VD, assert that getPropertyChangeWillBeConsumed
returns FALSE. This is due to the behavior change that we do not
set it unless the VD is dirty.
- set the bounds, so our onscreen check works.
- Add another test for prepareListAndChildren, which puts VDs offscreen.
Change-Id: Iae0a07adcf58e7884e0854720de644e7b2faf2bf
Currently only supported in the EGL path.
Vulkan support Coming Soon
Bug: 127822449
Test: trace of hwuimacro
Change-Id: Iac2b039e11d964aab5b8ca1bdf2a5430b187e2ea
Replace it with a newer, fancier, WorkQueue-inspired
one that's just a global common thread pool.
Test: hwuiunit passes
Change-Id: Ib5d03104a08bbac9a4ec67a1bfc0db2b35d6700f
Also adds a 'good' bar at 80% threshold
and a 'bad' bar at 150% threshold (where triple
buffering would start to get iffy)
Fixes: 127371028
Test: eyeball'd it
Change-Id: I109cd293f87f8cb9c8f1e66d49fb8fb2188b0bec
This reverts commit dd08224e4d.
Bug: 120097554
Bug: 120342364
Bug: 120421537
Test: verified emulator can boot OK, with no log warnings
Test: verified keyboard works in landscape
Change-Id: Iae2cd1140d81755b126adfb21ec885181f3bc6b6
You won't believe this one weird trick to make
EGL_BAD_ALLOC errors never happen again! libgui
hates it!
Bug: way too many to list
Test: none - there probably should be, though
Change-Id: I8e7fc3e584f90c01e0fd932497604a1d93710ba6
Split out the View/ViewRootImpl bits from the
hardware rendering bits.
Create a potential public API surface for
hardware rendering
Bug: 112709971
Test: builds & boots
Change-Id: I9e6f44b07a170574a905f42338282c4bb7e95f56
Reason for partially revert: Step back a little bit to think about how we want
to move the view system forward.
This patch removes the legacy mode we have added, as well as move color type
and color space of the surface to SkiaPipeline.
BUG: 111436479
BUG: 113530681
Test: Build, flash, boot and check dumpsys SurfaceFlinger
Change-Id: I1ab7b88d41347284d87649618e5a15f5c6bcf8fe
Previously, HWUI always produces SRGB buffers. We introduced new APIs for
SurfaceFlinger, a.k.a. the composer service to return to composition preference
for data space, and pixel format. This patch makes HWUI query composition
preference from composer service, and creates the corresponding EGL surface
with the correct attributes.
In legacy mode, HWUI will take the pixel value from source color space, and
interpret it as pixel value in destination color space.
BUG: 111436479
BUG: 113530681
Test: Build, flash, boot and check dumpsys SurfaceFlinger
Change-Id: I64562d5ea6f653076c8b448feb56b5e0624bc81c
Implement readback from Surface, TextureView and HW Bitmap
for Vulkan pipeline by wrapping the graphics buffer in an SkImage.
Refactor both Vulkan and GL readback to use common code.
TextureView readback is moved from IRenderPipeline interface to
Readback class. Refactor all 3 readback flows to use common
implementation.
Test: Passed all view, uirendering and graphics CTS tests with GL
Test: Passed many CTS test with Vulkan, that require readback
Bug: 113673613
Change-Id: Ifbfd8170a5401f87a709b4b1b9fa058e8e11768d
Fix an issue with incorrect texture matrix, when there is 90/270 rotation.
This happened, because after refactoring SkImage has buffer width/height,
instead of layer width/height.
This reverts commit a683eb3945.
Bug: 113673613
Test: Ran TextureView CTS and lensblur in camera app
Change-Id: If3bcf9cd5195de09fd67a753708568a8e3ca3a9a
Render TextureView as hardware bitmaps, instead of GL textures.
Cache SkImage for each observed GraphicBuffer, which is faster
even for GL.
Implement C++ SurfaceTexture, which allows Java SurfaceTexture
to be used with Vulkan HWUI render thread and application GL.
threads. Delete GLLayer and VkLayer classes and texture code
from old HWUI pipeline.
Test: Ran skiagl and skiavk pipeline with a TextureView app.
Test: TextureView CTS tests pass for GL pipeline.
Test: Ran Android NDK Native codec sample app.
Change-Id: Idc94f864ce2d34fd6ceff4be4fc7d3327e99879c
