* Add a DimmingStage to the client target configuration reported by HWC,
which may be used for some vendor color modes to configure when
dimming should occur during client composition
* Communicate the display brightness in nits, so that HWC
implementations do not need to parse the per-display xml to map from
display brightness to nits. Furthermore, this is more plausible for
extensibility for external displays.
Bug: 220396224
Bug: 218954037
Test: builds
Change-Id: I5675c16f0895f9958e3bee3ee4c85df8937ecdb7
An earlier patch replaced white point nits with a per-layer brightness.
This patch does the same for providing the brightness space of the
client target relative to the display brightness.
Bug: 217961164
Test: builds, boots
Change-Id: I1be65f7c511fefa239305e0735637126a1cd6622
Merged-In: I1be65f7c511fefa239305e0735637126a1cd6622
Check for support via getDisplayDecorationSupport. If it is supported,
verify that we can allocate a buffer of the correct format and that
Composition.DISPLAY_DECORATION can be used without error. If it's not
supported, expect an error of EX_UNSUPPORTED_OPERATION.
Bug: 209458568
Fixes: 209458568
Test: atest VtsHalGraphicsComposer3_TargetTest (this)
Change-Id: I0a62a0ce3f8fd3a2d6088f94ce1ad0840d9c6faa
When we have a fence created with GraphicBuffer
We don't need to wait on it in VTS tests,
the composer should be able to handle the
fence in this case
BUG: 219589185
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I688d695bb562dd1fe86cdceb642e746cbafe8b30
For multi-display order of the hotplug matters, and vector
maintains the order that we need for the multi-display.
see: ag/1921760 for partner cl on HIDL.
BUG: 210920960
BUG: 209409863
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I4f9a86413f20c860fc0bc3850a14335d62de881a
Client composition test was broken because there was a buffer
created at the test creation and this is not how
this test works for client composition trigger.
And for the fence that was used is different from what it
needs previously fence was acquired through gralloc but now
it's internal to the Graphic Buffer and we no longer
have to wait for the fence to signal it's taken care
by the graphicBuffer->unlock call, the unlock call waits
for us.
See for unlock https://source.corp.google.com/android/frameworks/native/libs/ui/GraphicBufferMapper.cpp;rcl=HEAD;l=139
BUG: 216170021
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I8fa25d8910a4e2b1df2f0e2270445a658e3b1a39
HIDL tests don't have any buffer created when test execution starts and with GraphicBuffer that wasn't the case.
We only initialize the buffer when required the way gralloc implementation did.
BUG: 199413815
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I3aeec5b00e30e636de2e58cf7e6ced5539ae19b6
There are several reasons for limiting the notion of white point nits in
the composer interface:
1. Some KMS apis exposed by drivers only expose a notion of a per-plane
brightness. While these are non-mobile drivers, e.g., nouveau, this
does indicate that white point is not directly going to be understood
by typical hardware
2. Changing the brightness without requiring a frame update introduces
implicit state in composer. If the brightness and white point nits
for a set of SDR layers are 200 nits, and the brightness changes to
205 nits to respond to ambient conditions, then composer must not dim
the layers, and in fact DisplayManager will tell SurfaceFlinger that
the SDR white point will be 205 nits. But SurfaceFlinger will not
tell composer that the SDR white point changed as that would
otherwise introduce a re-composition cycle, meaning that
HW Composer must track somehow that the layer white point changed
without a corresponding change on the layer data structure, which is
confusing.
3. It's poorly defined what the dimming ratio should be if
SurfaceFlinger provides the following inputs: Layer A has a white
point of 200 nits, Layer B has a white point of 400 nits, and display
brightness is 300 nits. Current implementations may clamp the
brightness of Layer B to be 300 nits and dim layer A by 2/3s, but
there is an equally valid interpretation which is just dim Layer A to
be 50% of Layer B's brightness.
4. The problem indicated by (2) and (3) suggests that layer white point is
really an up-stack concept, that SurfaceFlinger can be aware of for
properly computing the dimming ratios it can send to composer, but
the composer hal shouldn't really be speaking in terms of nits.
Note that this patch does not yet change the interface for
ClientTargetWithNits, which may be done in a follow-up patch.
Bug: 217961164
Test: builds, boots
Change-Id: I4a1b4e8c300d22599a5683bd44b7b8afa9a29425
In case a user fails to call setDimensions() prior to several
other methods on this class, we at least want to have consistent
behavior that's easy to debug.
Test: TreeHugger
BUG: 199413815
Change-Id: Iee7681ed855cf6ebf9e436e0d2864e0fc9cb4025
The test is removed from aosp so won't be required in HWC3.0
BUG: 200184776
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I21f0ccf94d1d92a730a41b579bf111fa0553f609
Makes corrections to the code such as checking EXPECT_TRUE, ASSERT_TRUE in certain cases, test name corrections, consolidate layer configurations.
See comments in ag/16465080 for change refrences.
BUG: 199413815
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I73c4621d8c525fdfbcd301d78a1ad97c142b2864
A wrapper helps with implementation details of IComposerClient and
improves re-usability with readback and base tests.
Changes from the original wrapper (ag/16578214):
Changed the createClient to return ScopedAStatus
as it helps more with the singleton test.
And updated getDisplays to go through all the configurations
and save it in the DisplayResources as we use
the properties like width, height, vsync later
in the test and we avoid querying to the composerClient for this.
With readback tests updated setReadbackBuffer method to use
const native_handle_t* buffer instead of const native_handle* buffer
BUG: 199413815
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: Ia33c9489bf15a52edcae55a91ec56666df3726c1
Different hardware may prefer different formats for DISPLAY_DECORATION
layers. Add a new struct with two enums: one for the format, and one for
how to interpret the alpha. These two properties are orthogonal.
Add IComposerClient.getDisplayDecorationSupport for retrieving the
struct. I0c2fbc78cf5e4e97cede4762699a1ecc98620919 demonstrates a sample
implementation.
Remove the DisplayCapability.DISPLAY_DECORATION, which is replaced by
this.
Bug: 216644902
Test: manual
Change-Id: I27f119f927b23052c5fd8f068cbca75338fe7b91
This reverts commit 634340d6a8.
Reason for revert: No one is using it, I will make a cl with usage of the wrapper along side this and reupload it.
Change-Id: I3ff3427fc9e81d53fd47a082ab027441f5dc8a66
A wrapper helps with implementation details of IComposerClient and
improves re-usability with readback and base tests.
BUG: 199413815
Test: atest VtsHalGraphicsComposer3_TargetTest
Change-Id: I627748d1c779eaacf92ff6fb5ccf3138a3024681
Remove the word asynchronous and insteam explain
that the cursor position can be updated without a validate/present.
Change-Id: Ia01bc097bc335779dba4e06577c285ca133dfd02
Fixes: 209711877
Test: make
Adds a Capability param as BUFFER_AHEAD to check if
BufferAhead is supported.
Test: Device boots after these changes.
BUG: 198186628
Change-Id: Ic0f3e602635a2519f4e90c546dfca1af985116f9
Stride must be retrieved from the graphic buffer immediately prior to
CPU access. The previous code was tripped up by an error-prone situation
where stride was never initialized, causing UB.
Also this patch sneaks in a couple of const modifiers for methods that
pass-by-ref.
Bug: 213493262
Test: VtsHalGraphicsComposer3_TargetTest
Change-Id: Ic601cca128b71d36876ed25bd55140b02cb0ad0f
If there's no readback buffer, then the status is NOT ok(), and the
returned fence must be -1.
Bug: 213493262
Test: VtsHalGraphicsComposer3_TargetTest
Change-Id: Iff1b6e04126bfcc748f0f8d173f63244bfd83be4
Add new functions to control display idle timer
- Adding an API to r configure display idle timer.
- Extend DisplayCapability to query display support idle timer.
- Adding an callback for notify client the display is idle, the refresh
rate changed to a lower setting to preserve power and vsync cadence
changed. The client can enable vsync callback to learn the new vsync
cadence before sending a new frame.
Bug: 194068871
Bug: 198808492
Test: build, vts
Change-Id: I47649d222022922926a05f2d700ca798f8f48b35
* Tests exercising that a bad readback buffer is sent should actually
use a native_handle from a deallocated buffer rather than a valid
handle, since that would still work
* Creating a readback buffer should not require passing in a buffer at
construction. Instead ReadbackBuffer should always allocate its buffer
internally whenever a new buffer needs to be generated.
* Passing the readback buffer's fence to lockAsync should be a dup'd fd,
as lockAsync takes ownership of the fence
Bug: 213493262
Test: VtsHalGraphicsComposer3_TargetTest
Change-Id: I73396f982cf9a67f0dd81eaed509cf9cc7b03314
Dimming is only verified if setting display brightness is supported. The
test then does:
* Sets the display brightness to max, which is the same nit level that
is found in the display xml file
* Dims a layer by 10%
Bug: 213493262
Test: builds
Change-Id: I3b552682276be26fefd7ae5586ba42af0e574e31
No readback tests as of yet, as that will necessitate a new api to get
the max display brightness, so that we can max out the brightness -> use
well-defined dimming ratios
Bug: 213493262
Test: builds
Change-Id: I0166dc67f4c52919c02ab7abbf47701e6be3d983
