* Move HintSessionWrapper into its own file
* Refactor HintSessionWrapper initialization and lifecycle
* Move HintSessionWrapper instance into CanvasContext
* Have HintSessionWrapper do its own validity checks
* Have HintSessionWrapper implicitly initialize when called
Bug: b/260267128
Bug: b/260224719
Test: manual
Change-Id: I69f89bb5ef601d1b5d9fca1a5da62256f2556da8
Sending HWUI ADPF CPU hints when the frame is not drawn can undercount
the actual workload duration and lead to inadequate boosting. This
patch prevents this by not sending hints when the frame is not drawn.
Bug: b/239677101
Test: manual
Change-Id: Ia308f59740c83977df02dff6d9f9289f2b8d439e
Should be a mostly no-op but adds a handful of new options
* Adds a 'MemoryPolicy' to mostly consolidate memory settings
* Moves trim handling into HWUI proper
* Adds settings for UI hidden & context destruction that's not
dependent on TRIM signals
* Treats persistent process the same as system_server
* Tweaks HardwareBitmapUploader timeout to reduce churn
Bug: 245565051
Test: builds & boots
Change-Id: I1f1b3db884ef7fa45ff2556436464a99440b998e
In the Vulkan pipeline, the GPU start time was measured to be when
swapBuffers starts. But the command queue has already been submitted at
this point, which means that the GPU work would already have begun. This
means that it's possible to measure a negative time since for very light
GPU workloads, the GPU fence can fire prior to CPU making it to
swapBuffers().
To compensate, instead measure from after skia completed submitting GPU
commands, until the GPU fence fires. Since it is theoretically possible
for GPU work to complete if the render thread gets descheduled
immediately after submitting the GPU, we also add some clamping to
ensure that the duration from command submission -> completion of GPU
work is nonnegative.
Bug: 230713131
Test: atest android.view.cts.FrameMetricsListenerTest
Change-Id: Ia30b7732eaab71e4e29766f788d5cd94ec63c38a
The forceDraw flag in HardwareRenderer will ensure a frame is drawn when
requested even if it would end up drawing multiple frames in a single
vsync.
This is to help blast sync when we want to synchronize the
buffer. We want to make sure we are guaranteed a callback since we don't
want to wait for retries, especially in the case when trying to synchronize
multiple buffers.
There was already a global flag to handle this, but would use the flag
for all draws. This new function is set per draw so once a frame is
drawn it's unset. The global flag was only used for tests so updated the
test to set the flag before every draw and deleted the global property.
Test: Underlying code was in place. This is just piping a new setter. No
usages yet.
Test: TestSceneRunner
Bug: 200284684
Change-Id: Ie1c9950cabb7331cfed1721564a51a1a15cd1624
* changes:
Get rid of unused ASurfaceControl in frame metrics listener callback
Set uninitialized frame number to 0
Make frame number type consistent
Add unit tests for FrameMetricsReporter
Update JankTrackerTests to reflect API changes
Pass surface control id to callback to accurately identify surface metrics belongs to
Stop reporting frame stats from frames completed before observer was attached
Now that the frame number is a uint instead of an int it's clearer if the uninitailized value is 0. Frame numbers start from 1 so 0 doesn't collide with any valid frame numbers. It also means comparing frame numbers to check if the frame is newer always works even when comparing to the uninitialized frame number.
Test: N/A
Change-Id: I0901ffeaefbdd6378e59d8858a06de92b0104cd9
The frame numbers are ultimately stored as uint64_t so to avoid any implementation dependent casting we make sure all direct and indirect references to this value are stored as uint64_t
Test: N/A
Change-Id: I080c94b1007f2e884cf0fa3583bf51badc913a05
Avoid getting the wrong frame info when duplicate frame numbers are found in the ring buffer. Will ensure there isn't a mismatch in the metrics data reported.
Test: Existing tests
Bug: 197515602
Change-Id: Iff9ba01f575f94e5a9872ee48c0dd1e5067880c3
The previous code waited for a frame complete callback from hwui before
notifying WMS that a draw had occurred. However, frame complete callback
was not guaranteed to get called since it only invokes a callback if a
draw actually occured. VRI really needs a signal that RT has completed,
since it just needs to know that a draw was possible so it can notify
WMS that the RT completed its pass.
Instead, rename frameCompleteCallback to frameCommitCallback since that
API is exposed to a public API when a frame was actually drawn.
Create a new callback, frameCompleteCallback, that is invoked when the
draw has completed, regardless if a frame was actually drawn.
When the frameCompleteCallback is invoked, VRI can check to see if a new
frame actually drew. VRI can call into BBQ to see if the frame acquired
matches the frame that was attempted to draw. If so, VRI can wait on a
transaction callback. If not, it can allow VRI to continue. In either case,
it will notify WMS that the draw has finished.
Test: Split over and over
Bug: 195262673
Change-Id: I24dd19ab2746be3fc33e597761abf8c5249f8b5b
Bug: 195081399
Test: manual + dumpsys gfxinfo
CanvasContext::draw has some conditions under which it will not draw. We
still need to flushAndSubmit to free scratch memory if there were
texture uploads this frame.
Change-Id: I60d230cd36bed310a3155e4f1ca647c99c13c4cd
This feature is intended for use in headless test environments,
where disabling drawing has minimal compatibility impact and can
save significant resource usage when running UI tests.
This feature is controlled via a new system property
'debug.hwui.drawing_enabled', and a HardwareRenderer public API that can
override the property value on a per-process basis.
Bug: 188559292
Test: boot emulator, atest CtsGraphicsTestCases
Change-Id: I93c21e680382e03342f235dbf58bd7a5c8a6f767
This field actually requires a special lock, mFrameMetricsReporterMutex.
But there isn't a GUARDED_BY annotation for it. And even if there was,
the compiler feature of -Wthread-safety was not active in this code, so
this error would not have been caught.
To fix this, enable the compiler annotation and add GUARDED_BY
annotation to mFrameMetricsReporter.
And finally, use this lock to properly protect this field.
Bug: 192330836
Test: atest hwui_unit_tests
Change-Id: I76950bfa01bbd7ccdc54c4e8c114430b5aeddf1a
Disable end-to-end input latency metric to reduce the
number of JNI WeakGlobalRef acquires per-frame.
Bug: 189738006
Test: Observe trace section no longer appears in calculator
Change-Id: I062dad8e7cec93cca5599e51d2b0b1a1b6e39a07
When the root surface control changes, the parent of the Webview surface control must be changed.
Bug: 191900007
Test: switch app, or turn on/off screen
Change-Id: Ib7b34133776fd5944afc638bfee471afe6a17349
A native transaction passed from webview is sent back to native side, so java side does not manage the life cycle of the transaction.
Bug: 191414767
Test: Play a video, switch to another app, wait for 10 seconds
Change-Id: I013052c202b445438d6cb6497f5f9a2fc22a2b85
Set the root surface control transparent, and set the z order of the newly created child surface control to -1.
A new callback is needed to update root surface control in sync from Java side.
Bug: 186750329
Test: use latest APKs from Webview team, play a video, mini toolbar should be visible
Change-Id: I0b37ee8f83fd2b41ff4f2856dbadd31ff6170baf
SurfaceFlinger will set the maxAcquiredBufferCount on the buffer queue
to account for the extra buffers needed, depends on the refresh rate
and app/sf durations, and there is no needed for any addional buffer
allocation is hwui.
Test: TBD
Bug: 188553729
Change-Id: Ic441a6feb15f6b084d45c9e538b11b7b24e36a0e
Added call to Skia's performDeferredCleanup method
to free resources that were not referenced within
the last 100 frames or 10 seconds whichever is
furthest away
Bug: 188450217
Test: manual
Change-Id: I3f37e1b5bd01330dbbc2da4a84b1259d56be2768
Since onSurfaceStatsAvailable gets called on binder-thread, we
need to ensure that instance doesn't get released while
onSurfaceStatsAvailable is calling reportFrameMetrics.
Test: Boots
Bug: 188934435
Change-Id: Iafe582d6fe4087a3c4274ee39a2803abaa363fd2
Subtract out time spent blocked in dequeueBuffer. Also subtract out any
time UI spents waiting on render thread while render thread is blocked
dequeueBuffer, though this calculation is fairly crude.
Test: Checked bouncyball returns reasonable-ish numbers even when HWUI
is ahead of surface flinger
Bug: 187556381
Change-Id: I368c446d93990ff8b7b645e30509405ba799c79c
The callback needs to set to null when thread renderer is destroyed.
Bug: 187419942
Bug: 186869429
Test: blaze test --test_strategy=local --test_arg=--device_broker_type=LOCAL_ADB_SERVER //javatests/com/google/android/testing/elizabot/internal/sanity/subscriptionleak:SubscriptionLeakTest_generic_phone_google_31_x86
Change-Id: Ic80c58f102ee5f21830542030021828f6231cc37
The basic idea is to create a child surface control from the root surface control passed from ViewRootImpl to the render thread.
Transactions are sent back to the java layer to get merged.
In case of offscreen layers, SurfaceControl must be disabled.
This new feature is disabled for Vulkan at the moment, a new CL will be used to enable the support.
Bug: 173671170
Test: manual, webview apks
Change-Id: I119405d13eca3c59fd3ec78e50dc7739f78411d4
On skiavk the flush to GPU happens sooner than it does on skiagl,
which was resulting in false-positive warning logs. Adjust the
condition to avoid this, so it should now only log if the GPU
fence time is significantly incorrect.
Fixes: 180488606
Test: doesn't log on skiavk
Change-Id: Ie6546b788663b5d863fb064f55245bead4eb0160
- Use GPU finish time as well as actual deadline to determine jank
rate.
- Use dynamic interval to adjust for 60/90hz switching
- Move frame metrics reporting into JankTracker to adjust the
deadline communicated to the app when in stuffing scenario.
- Adjust double-stuffing detection to be a bit more readable.
Test: GraphicsStatsValidationTest.java
Test: adb shell dumpsys gfxinfo
Test: FrameMetricsListenerTest
Test: Log output of FrameMetricsObserver
Bug: 169858044
Change-Id: I3a6b8ed163e2cf9cf2b67667110340ebe35f98a1
On Vulkan part of the abandoned check also checks for device lost.
Following what we currently do in GL, we now crash if we are ever
in this state.
Test: local build
Bug: 183722403
Change-Id: I2a41005fd6892e32a8b23dacbc0170d5542fe072
Currently, it is possible to notify frame metrics observers that
registered to receive present time about the same frame twice.
This is wasteful, but also requires the metrics listener to guard
against duplicates. See https://b.corp.google.com/issues/169866723#comment23
To avoid this issue, only notify about metrics with present time just
before the frame is advanced to the next position. This ensures that we
are always notifying about the last frame that's about to go away.
Bug: 169866723
Test: added a crash to LatencyTracker for the case where we are notified
about non-existent record. The record disappears once it's complete. So
the first call was completing the record, and the second call was
causing a crash. Now, the crash is gone with this patch.
Change-Id: Ia247b4a0010b8ecb016ac66e44e211ec4f40bf8f
To measure end-to-end touch latency, we need to report the actual
present time to ViewRootImpl. ViewRootImpl, in turn, will report this
information to InputDispatcher. Finally, InputDispatcher will combine
all known information for a specific input event, and will report this
data to westworld.
In another patch, we will add a new call, 'reportLatencyInfo', to
InputPublisher. This call will allow the app to send this latency data
to InputDispatcher.
Bug: 169866723
Test: printed the input event present times inside ViewRootImpl
Change-Id: Ibd3a2cfeb1a340eb15cd2165071df1f8589634af
Present time is available to apps, but is not currently stored inside
FrameInfo. In this CL, we add a location for storing the present time inside
FrameInfo.
Currently, the metrics reporter is triggered after the buffer is sent to
SurfaceFlinger. That means, metrics for anything that happens after gpu
draw completes are currently not available.
In a future CL, we will populate the 'presentTime' field.
In a future CL, we will register a metrics observer inside ViewRootImpl
in order to send this data to InputDispatcher. This will enable
end-to-end touch latency metrics collection.
Bug: 169866723
Test: printed present time inside FrameTimeline.cpp (surfaceflinger
side) and compared to the values printed inside a metrics observer
registered in ViewRootImpl (not done in this CL)
Test: atest ViewFrameInfoTest
Change-Id: I7d0f8c5d7b5a2572abdc4e107123e8938a36f582
- Add SurfaceStatsCallback to TransactionCompletedListener
- Register a callback in RenderProxy to be called when we have
surface stats from SF via the BLAST callback.
- Instead of finishing a frame for frame metrics reporting
immediately, wait until BLAST callback fires, note GPU completion
time and finish frame.
- Expose GPU_COMPLETION in FrameMetrics
- Modify TOTAL_DURATION to also include GPU_COMPLETION
Test: FrameMetricsListenerTest
Fixes: 171046219
Change-Id: I16fa1d80cfc4e7a5527c18fec7e885409f17ee4d
add a method, setSurfaceControl, for java layer to pass surface control to the render thread
Bug: 173671170
Test: call setSurfaceControl method in ViewRootImpl.java
Change-Id: I886a79c377938f19cf38b9058f2bec64e1439000
As HWUI uses vsync id to send buffers to SF, SF applies the right
amount of render ahead by latching the buffers at the configured time.
Test: launch an app and observe systrace
Bug: 178148035
Change-Id: Ifd3e1a2971aad0a085cb35d33e950194046aa634
0 and -1 both meant default when 0 should mean
0 and -1 should mean default
Test: manual
Fixes: 179290765
Change-Id: Ia9aa5e3d83757282bfff776e083d6b3d7d29e9c0
The application receives input events and produces graphic buffers in
response. We are interested in measuring the total time that the input
event takes to process, from the moment it's received on the device, to
the moment that the image is displayed on the screen.
To do this, we need to understand which input event produced a specific
buffer. In this CL, we are sending the input event id from FrameInfo to
SurfaceFlinger. This event id will later be used to identify a specific
frame, and provide the frame timing information to inputflinger.
Inputflinger will be able to use this information to reconstruct the
complete event timeline, and record metrics on the duration of each
input / graphics processing stage. This will allow us to optimize
end-to-end touch latency.
In the current CL, we are using 'NewestInputEvent' as the inputEventId.
That's not quite correct. Today, this field contains the timestamp of
the input event. Therefore, we will simply pass the truncated
timestamp instead of proper input event id. We will fix this in a
separate CL by providing input event id to FrameInfo.
Design doc: https://docs.google.com/document/d/1G3bLaZYSmbe6AKcL-6ZChvrw_B_LXEz29Z6Ed9QoYXY/edit#
Bug: 169866723
Test: printed input event id at the site of SurfaceFrame creation in
FrameTimeline.cpp
Change-Id: Ia90337bb2f000e9c93a4db04d9dd6ea9ea153520
