This reverts commit 56866118.

Reason for revert: b/294998861

Change-Id: I8616665f69be8e827529ed3a76b337606b900dfc

This improves performance by moving allocateHelper to a different
thread, which is launched early so that the allocation is ready when
it is needed. In addition, each display will only need at most one
additional texture per frame, so replace the idea of a minimum pool
with making sure there is either at least one texture in the pool
or a future which will generate one.

Bug: 256184546
Test: atest libcompositionengine_test
Change-Id: Icb92a85c07b9f2911d1365aa9bbc8d0750a8c116
Merged-In: Icb92a85c07b9f2911d1365aa9bbc8d0750a8c116

cache buffer rendering supposes to utilize the cpu gap after current
frame finised and before the next frame. But it could take long and
delay the next frame start time and cause jank eventually.
Sending CPU LOAD UP hint to prevent stealing time from the next frame.

Bug: 285991802
Test: PtsUiBench QHD
Change-Id: Id619a8166ae253fef815a25a35fc4b82066144a9

with null

Fixes a regression introduced in T which ignores a setBuffer call
with a null buffer. The expected behavior should be to release the
currently presented buffer from surfaceflinger. The subsequent frame
will not present this layer so the region behind the layer will be
composited instead.

Bug: 241271897
Test: presubmit
Change-Id: Ie06025c59c58cc75a267b783729996a3cbceef45

When the next expected presentation time is not allocated for
the previous frame, SF needs to wait before submitting it to HWC
to prevent early presentation.

Bug: 273419557
Test: tested by partner
Test: atest ASurfaceControlTest
Change-Id: Ia83755bbc71968369fd9176944b9a85135d8a04a

Bug: 267350616
Test: builds
Change-Id: I75bb485f01ae2358c77133767f01cb63f8eda6f1

- disable caching if hdr/sdr ratio > 1.01f, which we uses the extended
  brightness api.

Bug: 273624686
Test: play HDR video in silkFX and make sure no overdimming obeserved;
atest libcompositionengine_test

Change-Id: Iae53c5928e4f91b4b73f9e4ee9dd7bb052e86090

- separate color and bgcolor in layer state so we don't set a color
 on a layer without a buffer
 - avoid using legacy layer state when latching a buffer
 - fix callback issue where invisible or offscreen layers were not invoking
 the callbacks
 - pass in layer snapshot for trusted presentation state
 - fix a screenshot issue where the root layer was skipped
 - pass in framerate changes to layer history

Test: presubmit
Test: atest SurfaceFlinger_test with new fe
Bug: 238781169

Change-Id: Id9ff8a036dc283e21a05985c1c01ebd070b1df24

Some system layers may need to be excluded from surfaceflinger caching
for a variety of reasons, including power or image quality
considerations. This provides a mechanism for those system layers to
never be cached on the GPU.

Bug: 259311918
Test: libcompositionengine_test
Test: run test app and inspect planner state
Change-Id: I35418ad5a41cb2546e3b27b0af290bf3c31a0180

1. Ignore display overlays since we don't want screen decorations
   included in the occlusion

2. Handle occluded region as separate Rects to ensure that disconnected
   Rects in a Region are not considered occluding in the disconnected
   area.

Test: LayerTrustedPresentationListenerTest
Bug: 256993331
Change-Id: Ib0a4b850e2aafb42e206b8728fcc9b6013171f3f

* Removes redundant logic for determining whether or not SurfaceFlinger should request texture filtering.
* Fixes an issue where ScreenCaptureOutput was using incorrect values for its layer stack projection space.

Bug: 238348307
Test: atest TextureFilteringTest

Change-Id: I4ece87321ee73d10b86a2c01d53717d337c11926

Bug: 241001465
Test: manual currently, flagged off
Change-Id: I707281b9acaf6ea218f1d9ce888fc7cdbf1cf7c9

This allows the memory for the active buffer to be freed as soon as
possible by purging it from HWC cache as soon as the next buffer is sent
to the layer.

Bug: 258196272
Test: atest OutputLayerUncacheBufferTest
Change-Id: I96c24390de5757ac99b369119e9ba031afb0b042

When an app discards graphic buffers, for example when a MediaCodec
disconnects from a surface, those buffers will be uncached and removed
from HWC buffer slots. The active buffer, however, should still remain
active until the next buffer is queued up.

Bug: 262037933
Bug: 258196272
Test: atest OutputLayerUncacheBufferTest
Test: atest VtsHalGraphicsComposer3_TargetTest
Test: atest VtsHalGraphicsComposerV2_2TargetTest
Change-Id: I7c4eefb17e8bad694d698f9ad6d1d289f4af8d2c

When buffers are discarded by SurfaceFlinger clients, the memory should
be free'd immediately. This includes clearing references to them from
within the implementation of Composer HAL, specifically in cache slots
associated with the layer.

Since there is no HAL API to clear the slot directly, prior to writing
any other buffer to the layer, we set the layer's buffer multiple times
to a placeholder buffer, using each of the slot numbers that need to be
cleared. This replaces the reference to the client-discarded buffer
inside Composer HAL with a reference to the placeholder buffer, causing
the reference count to drop, allowing the buffer memory to be freed.

Bug: 258196272
Test: atest OutputLayerUncacheBufferTest
Change-Id: Id85482e8859490566f0eedbd8d8729c47a7349fb

Stop caching buffers inside SurfaceFlinger and remove the tight coupling
between SurfaceFlinger's ClientCache and the Hardware Composer cache.
This allows a better seperation of responsibility, where buffer cache
management is not split between HwcSlotGenerator and HwcBufferCache, but
is instead solely handled by HwcBufferCache.

Note that FramebufferSurface and VirtualDisplaySurface no longer use
HwcBufferCache, but instead use their own cache slot management that
is solely based on BufferQueue slot numbers.

Also do minor refactoring in FramebufferSurface to simplify code.

Bug: 258196272
Test: started and stopped multiple YouTube videos on adt4 and verified
no change in graphic buffer usage
Test: atest HwcBufferCacheTest
Test: atest OutputPrepareTest

Change-Id: Ica7955ab4bc70e3c70207390e36dff73a2fc4949

Bug: 233818658
Test: Manual
Change-Id: Iff92fc56c43aefdb4fd978e837969f8ddc9c24f4

On a multi-display device, it is helpful to know which display
particular methods are called for. Add information to traces specifying
the name and id of the display. The device still needs to specify
different names for displays (see b/254851304), but the traces will now
show the id, which is helpful until the device is updated, at which
point the names will be immediately helpful.

Output:
- Store a string that combines the readable name and the DisplayId. This
  saves allocation costs for the various places where we add this info
  to the trace.
- Print the name and id in present and postFramebuffer. Although
  postFramebuffer is called by present, future CLs will move it to a
  separate thread, and this allows tracking which call applies to which
  display.
DisplaySettings:
- Add a field with the display name and id. This allows SkiaRenderEngine
  to print them as well.
SkiaRenderEngine:
- Print the name and id in drawLayersInternal
- Replace the outdated text for the function name referring to the class
  as "SkiaGL" with a simple call to __func__.
Display:
- Print the name and id in chooseClientCompositionStrategy, which may
  also run in another thread.

Bug: 241285473
Test: manual; look at traces
Change-Id: I3081c4e4a7b5874139af6b5dd74a6a8ab0ad8cf7

Bug: 238643986
Test: presubmits
Test: go/wm-smoke

Change-Id: I6dbbcd5ce5070ec1d801dca55b5bb89fafe839cb

This CL allows CE instances without TimeStats. This will be used by the
short lived CE instances created in the upcoming renderScreenImpl
refactor.

Bug: 238643986
Test: presubmits
Change-Id: I7f996f3ce660b335f6228069aa04bba617806b68

This change allows multiple instances of CompositionEngine to use
the same RenderEngine instance which enables creating new instances of
CompositionEngine to run off the main thread. This will be used in the
refactor of SurfaceFlinger::renderScreenImpl to generate LayerSettings
via CompositionEngine.

Bug: 238643986
Test: presubmits

Change-Id: I0c8dc30e091578d0d77afb434a52038db6e82288

Make LayerFE a standalone class with no links back to Layer. Pass
LayerSnapshot from Layer into LayerFE before CompositionEngine::present
and back after.

Bug: 238781169
Test: go/wm-smoke
Test: presubmit

Change-Id: I5395fb717a931f88e2bf26395acd21e8b308961e

The current logic only checks if dirty is not empty, but doesn't account
for no content changing. Instead use the same logic that's in
beginFrame to compute mustRecompute and use that when determining if the
display should queue a buffer.

This fixes an issue where a Displays first few frames could be black
until the content is loaded.

Bug: 239888440
Test: First frame from screenrecord is no longer black
Change-Id: Ibb568302bf2d50db167a1f9ca76a0f37e2c65bf5
(cherry picked from commit 09fa1d66)

Adds LayerMetadata to LayerSnapshots. LayerMetadata is only populated
when CE enables the new feature flag and the feature flag will only be
enabled in ARC++'s CE.

This change is necessary to split LayerFE from the Layer class as
ARC++'s LayerFE subclass accesses LayerMetadata.

Bug: 238781169
Test: go/wm-smoke
Test: presubmit

Change-Id: Iae405707897eedf5af9025fe628c1cd60b596e10

First step in breaking CE dependencies with frontend.
FE should provide a snapshot of the state for CE to
consume. This cl updates the composition state
based on whether visible regions are dirty before
calling into composition engine.

Test: presubmit
Test: go/wm-smoke
Bug: 238781169
Change-Id: I6efd71cf66d9e1cae163c98ad4ce3097e09848af

After SurfaceFlinger receives the brightness commands, the commands are
sent to HWC until the next frame presents. However, no following frames
are present after the power mode is set to off, and HWC cannot receive
the last brightness commands.

This change forces SurfaceFlinger to flush the staged brightness to HWC
prior to execution of power off.

Bug: 239908529
Test: suspend and resume repeatedly
Change-Id: Ia8fb04399e757de16e06e6516d86bdfcfabd5a2e

The current logic only checks if dirty is not empty, but doesn't account
for no content changing. Instead use the same logic that's in
beginFrame to compute mustRecompute and use that when determining if the
display should queue a buffer.

This fixes an issue where a Displays first few frames could be black
until the content is loaded.

Bug: 239888440
Test: First frame from screenrecord is no longer black
Change-Id: Ibb568302bf2d50db167a1f9ca76a0f37e2c65bf5

Bug: b/232535621
Test: atest librenderengine_test
Test: atest SurfaceFlinger_test

Change-Id: If036ab847024751b8da97d6ed856db447db19f6f

the behavior in Output::composeSurfaces().

Bug: 240293363
Test: atest libcompositionengine_test. Fixed tests, added CachedSetTest::renderWhitePointNoColorTransform

Change-Id: Ic0317b34978c2ae8d5c057c0a39ed889b86b3da0
Merged-In: Ic0317b34978c2ae8d5c057c0a39ed889b86b3da0

Plumb through Output::getSkipColorTransform() into CachedSet, to match the behavior in Output::composeSurfaces().

Bug: 240293363
Test: atest libcompositionengine_test. Fixed tests, added CachedSetTest::renderWhitePointNoColorTransform

Change-Id: Ic0317b34978c2ae8d5c057c0a39ed889b86b3da0

Replace Layer::prepareClientCompositionList with
Layer::prepareClientComposition now that at most one layer is returned
for any call. Also changes OutputLayer::getOverrideCompositionList to
return an optional for the same reason.

Output::generateClientCompositionRequest is updated to use buffer ids
instead of strong pointers to buffers when determining whether or not
override settings are redundant. This is done to avoid duplicate checks
on whether or not overrideInfo has a non-null buffer.

Change-Id: I777f6ba8c3ca38ea31773e6fcbacb65fad287b03
Bug: b/188891810
Test: atest libcompositionengine_test

The corresponding backdoor has been removed, and dumpsys output has no
practical use.

Bug: 241285477
Test: m libcompositionengine_test
Change-Id: I1291b6164d5ef6efa9f94498f90206684dacb779

Bug: 241680050
Test: presubmit
Change-Id: I78a3effb33021b79d451f5e9f32a7d96226bdc60

Layers with SOLID_COLOR compostion type may cause fence leaks, and since these layers are not frequently used, skipping them should be ok.
Bug: 230073351
Test: manual, examine logs

Change-Id: I50c2488c4c5b891bf0415f1d4bd29a6e0c49be5b

ftl::yield, which lifts T to std::future<T>, incurs the cost of
allocating, ref counting, and locking the latter's shared state.

Consolidate the existing std::future extensions into ftl::Future,
and optimize ftl::yield by including static storage for T within.

Bug: 232436803
Test: simpleperf (-31% cycles in postFramebuffer)
Change-Id: I9a7ca7de17e7af10515de97d2f6a0dfa24e35d7a

Convert the unique_fd of RenderEngineResult (and futures thereof) into
sp<Fence> such that postFramebuffer does not duplicate release/present
fences.

Remove a few copies of shared futures/pointers with std::move.

Bug: 232436803
Test: simpleperf (-33% cycles in sys_dup)
Change-Id: Ia7c6c8333a712441f3612fb5c720ea2932799636

As of recently we have begun speculating on the composition
strategy in cases where we expect GL composition to be present.
This enables us to gain back main thread time by running
chooseCompositionStrategy (sometimes a long call to HWC)
in parallel with GL composition itself. Currently this is only
enabled when no geometry changes. Unfortunately this doesn't
help us a lot, since most of the cases where we have jank
and issues with SF runtime are around animations. In this
CL we try and expand the heuristic to cover more cases,
including app launch. Now we expect that if we are sending the
same output layers to HWC in the same order, and all we have
changed is geometry, then the composition strategy will remain
stable. We implement by hashing a subset of the values
sent to HWC and detecting changes.

Test: Existing tests pass. Play with phone lots, no mispredictions
Merged-In: I1f41e2cc67b1b55780e1bf6e837d766b4c186161
Bug: 230617082
Change-Id: I1f41e2cc67b1b55780e1bf6e837d766b4c186161

As of recently we have begun speculating on the composition
strategy in cases where we expect GL composition to be present.
This enables us to gain back main thread time by running
chooseCompositionStrategy (sometimes a long call to HWC)
in parallel with GL composition itself. Currently this is only
enabled when no geometry changes. Unfortunately this doesn't
help us a lot, since most of the cases where we have jank
and issues with SF runtime are around animations. In this
CL we try and expand the heuristic to cover more cases,
including app launch. Now we expect that if we are sending the
same output layers to HWC in the same order, and all we have
changed is geometry, then the composition strategy will remain
stable. We implement by hashing a subset of the values
sent to HWC and detecting changes.

Test: Existing tests pass. Play with phone lots, no mispredictions
Bug: 230617082
Change-Id: I1f41e2cc67b1b55780e1bf6e837d766b4c186161

Update the enableBorder API in SurfaceComposerClient so
that width and color can be specified for the border.
The information propagates through the pipe all the
way to the render engine so it can render the correct color
and width for the border

Test:go/wm-smoke.
Test: LayerBorder_test
Bug: 226529222

Change-Id: Id3ab853d5b4d6899a915f729b0d7be701cb5b167

Currently we issue hints at the end of the frame reporting what happened
that frame. However, this is often too late anyway and lacks critical
information not known until after the frame completes, like gpu timing.
By waiting until the next frame start and using those timings along with
information about gpu duration, the new expected present time, and the
new wakeup time, we can instead use those timings to predict the
duration of the upcoming frame and signal in advance of a frame drop.
This patch also changes how and where frame timing is done and measures
gpu duration so we can predict if a cpu speedup can help gpu composition
happen earlier, preventing a drop that way.

Bug: b/195990840
Test: manual

Change-Id: I5f0ae796198631fced503dce8e6c495885fb256a