summaryrefslogtreecommitdiff
path: root/libs/hwui/Properties.cpp
AgeCommit message (Collapse)Author
2025-02-11Calculate workload target using original frame deadline Melody Hsu
The target work duration in HWUI ADPF is sometimes incorrect, and is more often incorrect during buffer stuffing recovery due to the exposure of new jank. The issue causes CPU clock frequencies to decrease too much, and cause potential jank in subequent frames. The original frame deadline should be used instead in calculating the workload target deadline when triple buffered because even when the deadline is pushed later during buffer stuffing, the throughput requirements remain the same. Also fixes a misordering in FrameInfoNames that was inconsistent with FrameInfo and FrameMetrics. Bug: b/389939827 Test: atest JankTracker Test: presubmit, perfetto traces w/ and w/o buffer stuffing recovery Flag: com.android.graphics.hwui.flags.calc_workload_orig_deadline Change-Id: I2e3da6ee927d20fcbc9299ab5eda5732660d7246
2024-12-13render thread early preload optimizations “Shadman
1. preload eglContext when renderthread preload is called so setSurface work is not waiting for this work during app launch crit path. 2. Additionally queue GraphicBufferAllocator instance get, so allocateBuffer does not wait for initialization during app launch critical path. Impact: Large improvement in app launch latency for low core devices - ~70ms of improvement in our view based app launch - ~27ms of improvement in material3 compose base app launch Test: manual built and flashed Bug: 383612849 Flag: early_preload_gl_context Change-Id: I580e31de1d3a9878cf102dfdcad39bfd64bf2c53
2024-11-09Fix seaming issue in BitmapRegionDecoder for gainmaps Alec Mouri
Rounding discrepancies in the image codec infrastructure was allowing gainmaps to be decoded with a 1px border on the edge, which introduces visible seams when zooming into large images. Fix this by adjusting how computing the desired gainmap subregion size, by pretending we downsample first, then compute the subregion size based on that. Add a long paragraph explaining why we have to do this with an example, because it's not entirely intuitive. Also adjust how we poke at the gainmap resampling flag, since some CTS tests run before RenderThread starts up and intializes the properties. Bug: 371123308 Flag: com.android.graphics.hwui.flags.resample_gainmap_regions Test: Decoding in Photos and Files Test: GainmapTest Change-Id: Id608c5838f7890d0304b040e13d7281aa879a75c
2024-09-19[Vulkan] Query global priority support for queue creation Juston Li
Setting up a device queue with a global queue priority above the system default requires sufficient privileges (SYS_NICE) otherwise queue creation will fail. SurfaceFlinger and system_server have SYS_NICE and will set priority to REALTIME. SysUI and Launcher take that as an indication they should request HIGH for better preemption strategy. However for upstream DRM drivers, we currently lack a way to grant them granular access to HIGH but not RT. For long term, the right way to do this is via drm specific cgroup for scheduling controls however this is currently being worked on upstream. In the meantime, instead of fatally crashing, we can query the global priorities supported before queue creation and if its not supported, drop the priority request and log a warning. Note this requires vkGetPhysicalDeviceQueueFamilyProperties2 which is provided by Vulkan 1.1 while the global priority query can be safely added to the pNext chain and will be ignored if VK_EXT_global_priority_query isn't supported. Bug: 343986434 Flag: com.android.graphics.hwui.flags.query_global_priority Test: UI comes up debug.renderengine.backend=skiavkthreaded Change-Id: I662c9ce3f3724e87690e25d260ee010340451c53
2024-08-09Skip eglCreateContext telemetry for HWUI contexts Tom Murphy
We want eglCreateContext telemetry to give us insights into how developers use EGL. Including HWUI telemetry in this pollutes this telemetry. Skip recording of the creation of this context using the new Egl extension. Currently this only disables the passing of GpuStatsInfo::Stats::CREATED_GLES_CONTEXT to the gpu stats. Test: Checked the code path executed correctly with the flag enabled/disabled using logging Bug: 347911216 Flag: com.android.graphics.hwui.flags.skip_eglmanager_telemetry Change-Id: Ifcadd5af263f26388026bcd70ec8dc9c2914cd0a
2024-08-08Resample gainmaps during region decoding. Alec Mouri
Implicitly, region decoding is a cropping operation, which previously was introducing quality issues when decoding a gainmapped image. For instance: (a) Consider a 48x48 image with a 12x12 gainmap. (b) Consider decoding a 10x10 region of the image. The decoder must choose to either decode a 2x2 or 3x3 region of the gainmap, which does not match the 4x scale ratio of the source, so either we decode too much or too little of the image (c) When displaying the image, we bilinearly scale both the image and the gainmap to the destination, then apply the gainmap. But, because of (b), the gainmap is misaligned with the base image, introducing haloing artifacts. To fix this, we change (b) by always decoding a slightly larger region (in the examplar case, either a decode a 3x3 or 4x4 region), but retain information about the "logical" region we intended. Then, we resample from the "logical" region up to the decoded bounds. In the above example, this means that we decode a 3x3 or 4x4 bitmap, that holds the logical 2.5x2.5 region of the gainmap, ensuring that (0, 0) in the resulting bitmap maps to (0, 0) in the bitmap for the decoded region from the base image, so that the gainmap content lines up with the base image after the gainmap is upsampled during rendering. We do the actual resampling inside of the recycling allocator, since we sometimes perform a bitmap copy there anyways, so we can resample during the copy. Hide this behind a flag, since I broke decoding in about 5 different ways before settling on this. Bug: 352847821 Flag: com.android.graphics.hwui.flags.resample_gainmap_regions Test: Decoding works in Photos, Files, and modified SilkFX Change-Id: Ic21d44011858619273b11c20ee746614a1749a73
2024-07-29Merge "Use ro.hw_timeout_multiplier in Readback deadline" into main am: ↵ Jason Macnak
2eb7e9d8a8 Original change: https://android-review.googlesource.com/c/platform/frameworks/base/+/3198371 Change-Id: I0c06b96ac45a93d161f16c4b3b931fb3167feaf5 Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
2024-07-29Use ro.hw_timeout_multiplier in Readback deadline Jason Macnak
... to give SwiftShader some more time to complete. Bug: b/354222246 Test: CtsUiRenderingTestCases x100 Change-Id: I943249a84daa873508bddcfc532048ef71eb4129
2024-06-04Initialize GL even when HWUI is running Vulkan Leon Scroggins III
Even though more and more devices are running Vulkan, many apps still use GL, instead of or in addition to using HWUI. Initialize GL as part of the zygote step. This only happens once at boot - new app processes won't do this again, so it speeds up app startup time. And since it happens before forking the zygote, it starts off in shared memory. If an app never uses GL, it won't have to use any memory for this GL state. Use a new static method on Properties, rather than a static bool, as is done for several other properties, because this happens before loading other properties. Bug: 335172671 Test: manual verification via log statements Flag: initialize_gl_always Change-Id: I80b366121db0034131afb656ea774e37171f223b
2024-03-13Use HWUIProperties.sysprop in host builds Jerome Gaillard
It uses android-base properties now, so is compatible with host. Bug: 322360037 Test: build libhwui on host Change-Id: I6502fe3001aac3d95ea2cd2c8327b8fe7baae55a
2024-01-16Merge "Support fp16 in hwui for HDR" into main Alec Mouri
2024-01-11Add default shortcut for tracing Skia's GPU memory usage Nolan Scobie
Bug: 319164766 Test: comparing traces with debug.hwui.skia_tracing_enabled on and off Change-Id: Ie24e3b4705380ef143699f229469226df7b4913a
2023-11-08Merge "libhwui: Use Vulkan by default" into main am: 822a45e2e1 am: ↵ Leon Scroggins
f9a31c7297 am: 74c7d271c4 Original change: https://android-review.googlesource.com/c/platform/frameworks/base/+/2819917 Change-Id: If6f9cb0298f5a84a3b616f747afe21fd3c2e704f Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
2023-11-07libhwui: Use Vulkan by default Leon Scroggins III
Pixel has been using Vulkan in libhwui for several years now. We strongly recommended using it in the Android 14 (U) CDD, and we plan to make it MUST in Android 15 (V). Update the default to match. Bug: 293371537 Bug: 301558076 Test: CtsUiRenderingTestCases, CtsGraphicsTestCases Change-Id: I106381fadf6a686127cb6120afbaef53db5c4aa8
2023-11-03Support fp16 in hwui for HDR Alec Mouri
Picking sRGB primaries and transfer with extended range bit as the dataspace, because that's straightforward for EGL. Then, rather than dimming SDR content, we instead just render as normal so that we take advantage of the full bit depth Bug: 236745178 Test: builds Change-Id: Ica39a2c81be508effb97a8850b0c8e272c10a084
2023-10-27Implement automatic SV clipping John Reck
Clip SV to its ancestor clipping bounds. This enables Z-above SurfaceView + scrolling containers to work more naturally Replaces the hidden API of setEnableSurfaceClipping Fixes: 298621623 Test: Sample app Change-Id: Iaa862598e37065677f5ba163a5ac7a6fab2739ea
2023-05-25Revert "Turn on vk unit tests" Alec Mouri
This reverts commit aa3e498bbc9825b8c07a2094dc83e15b6bb47ffc. Reason for revert: Synchronization with WebView is hard. Also: * All currently shipping devices that turn Vulkan on for HWUI support at least two graphics queues, with the exception of emulators using swiftshader. * Using two graphics queues doesn't significantly regress RAM on those devices Bug: 280564832 Test: DailyMail app (cherry picked from https://googleplex-android-review.googlesource.com/q/commit:219997a1e7c07a70aa935f09dbff9fb53ed03cb7) Merged-In: Ica0c461af9df771292967ed10df63b2aa6902ae3 Change-Id: Ica0c461af9df771292967ed10df63b2aa6902ae3
2023-05-23Revert "Turn on vk unit tests" Alec Mouri
This reverts commit aa3e498bbc9825b8c07a2094dc83e15b6bb47ffc. Reason for revert: Synchronization with WebView is hard. Also: * All currently shipping devices that turn Vulkan on for HWUI support at least two graphics queues, with the exception of emulators using swiftshader. * Using two graphics queues doesn't significantly regress RAM on those devices Bug: 280564832 Test: DailyMail app Change-Id: Ica0c461af9df771292967ed10df63b2aa6902ae3
2023-03-23Merge "Fix hwui compile error" am: 9bbcd6e11e am: b867bee3da am: 0580b4fdf1 ↵ Leon Scroggins
am: 8adc586787 am: 61699386a1 Original change: https://android-review.googlesource.com/c/platform/frameworks/base/+/2495575 Change-Id: I6270a638338901c009909cba0fd222aa4585b2c1 Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
2023-03-23Merge "Fix hwui compile error" am: 9bbcd6e11e am: b867bee3da am: 0580b4fdf1 Leon Scroggins
Original change: https://android-review.googlesource.com/c/platform/frameworks/base/+/2495575 Change-Id: Ie5d699819c295b25ac8b2f9f579235b957140f71 Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
2023-03-17Fix hwui compile error tangcheng
set framework/base/libs/hwui/Debug.h DEBUG_INIT to 1 then compile libhwui, there is a error Bug: 274030026 Change-Id: Ic8857b0510c720b8108bde470a6a00e737c1f6fa Signed-off-by: tangcheng <tangcheng@xiaomi.com>
2023-03-15Merge "Add sysprop to control HDR headroom" into udc-dev am: 954c3d6d3e am: ↵ John Reck
008dc4a408 Original change: https://googleplex-android-review.googlesource.com/c/platform/frameworks/base/+/22087306 Change-Id: Ib11412242ecb2af89520045b2559899746f30460 Signed-off-by: Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com>
2023-03-15Add sysprop to control HDR headroom John Reck
To accelerate tuning of the value allow it to be sysprop driven Bug: 273624684 Test: change prop, verify silkfx headroom changes after re-launch Change-Id: Ibad662b7ba06bd821c7a3ce48c19f4c94c716b5e
2023-02-16Remove use of private SkBlurMask.h Kevin Lubick
This also makes a use of src/core/SkTraceEventCommon.h explicit. These are small steps toward removing Skia's src/core from the hwui include search path. Change-Id: I4fe843a4e44a675748bfe2cef329202eb4dab7bf Bug: skia:13983
2023-01-17Rename HWUI's Skia tracing sysprop and add new sysprop for Perfetto in Skia Nolan Scobie
Bug: 259248961 Test: none? Change-Id: Ieb6cc796e31ae5567206a07ae61391ac91ed488d
2022-12-06Disable HWUI adpf by default and enable in device.mk Matt Buckley
* Make HWUI disabled by default for vendors * Explicitly enable in supported devices * Consolidate flags for HWUI and SF Bug: b/261628396 Test: manual Change-Id: I2d6b53459e4f9643e6b0bf2ba81eb4db36fb94b4
2022-09-28Change how memory policy is configured John Reck
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
2022-05-13Merge "Revert "HWUI: Turn on Skia tracing by default"" into tm-dev TreeHugger Robot
2022-05-05Revert "HWUI: Turn on Skia tracing by default" Leon Scroggins
This reverts commit 7024d83459ec847b57527332d86fc12022b87227. Reason for revert: Performance regression Bug: 221813628 Fixes: 224677119 Change-Id: I276067cd8183b5b30914a78cbe48315ad7d7170d Test: perfetto
2022-03-22Add forceDrawNextFrame function for HardwareRenderer chaviw
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
2022-03-10HWUI: Turn on Skia tracing by default Leon Scroggins III
This can provide useful information when diagnosing jank. It also introduces a little bit of overhead, which is why we haven't turned it on before. That said, the overhead is only relevant during tracing. Bug: 221813628 Test: perfetto Change-Id: I311d902797e0b9ecf53631f2f9f2df6cf4860635
2021-10-08Merge "Update RenderEffectCache capability query" into sc-v2-dev am: ↵ Nader Jawad
58745e540e am: 6c8f89e428 Original change: https://googleplex-android-review.googlesource.com/c/platform/frameworks/base/+/16007299 Change-Id: I6c4f7dd465be57034b8f32aa4539c3a098330dfa
2021-10-08Update RenderEffectCache capability query Nader Jawad
Updated logic to determine if RenderEffectCache is supported to very both the vendor name as well as the corresponding driver version. Bug: 193145089 Test: Added test to EglManagerTests + RenderEffectCapabilityQueryTests Change-Id: Ia0fd66c40790b94566f495d2c4128c4988d406b6
2021-09-23Merge "Conditionally cache RenderEffect results" into sc-qpr1-dev am: ↵ Nader Jawad
69617682e4 am: 53290d8b19 am: 15fcba18e1 Original change: https://googleplex-android-review.googlesource.com/c/platform/frameworks/base/+/15882028 Change-Id: I9112e09456fb346e370446246c4f01a42ac63d43
2021-09-22Conditionally cache RenderEffect results Nader Jawad
Added property to conditionally cache SkImage instances with SkImageFilters applied. Some GL vendor implementations invoke Fence::waitForver without signalling, leading to ANRs throughout the system. Bug: 193145089 Test: manual Change-Id: I3e478b11b66205d96e9499a362f7412a1d6e952d
2021-08-17Merge "Turn on Webview Overlays" into sc-dev am: a43c6c4f2c am: 0d8ee15a18 Huihong Luo
Original change: https://googleplex-android-review.googlesource.com/c/platform/frameworks/base/+/15559504 Change-Id: I9438a92ab7df133bd77ae9e59c6aa82f8809535d
2021-08-16Turn on Webview Overlays Huihong Luo
Change the default value of the sysprop to be true. Bug: 173671170 Test: install webview apks from Vasiliy, make sure surface control is used. Change-Id: Idf495a2be3b2bdba4fdaa380a5dbe9fd96158a97
2021-07-16Lazy load Properties::isDrawingEnabled. Brett Chabot
Properties::isDrawingEnabled can unexpectedly return true even if the system property debug.hwui.drawing_enabled is false, if its called before RenderThread is created. This commit changes the underlying logic to lazy load the debug.hwui.drawing_enabled prop value. Bug: 193920054 Test: call HardwareRendererCompat.isDrawingEnabled()) from a new test Change-Id: Ib69577745c26595ed3964a4341086138dd5d34b0 (cherry picked from commit 8e1ab42f51c501949211746c46a02a7564ebf292)
2021-07-13Add ability to disable GL draws. Brett Chabot
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
2021-06-30Add a sys prop to control WebView Overlays support Huihong Luo
The boolean system property is named as "debug.hwui.webview_overlays_enabled" Bug: 192267127 Test: change the property value, check presence of Webview Surface Control Change-Id: I01e3e26282a5fa79aa504a6e49c5abe1a1c3ea02
2021-05-17Add support to toggle different shader Nader Jawad
behaviors Updated HWUI to toggle overscroll stretch implementation based on whether the device supports high end graphics or not Bug: 187718492 Test: manual Change-Id: I13a91a8861c07bec8af43268ba22d0f5b7060b4f
2021-05-11Add some more scenes, replace linear with uniform John Reck
Test: this Bug: 187718492 Change-Id: I171b5c15cd18645c2da1cffffa9635f83f8877e6
2021-05-10Updated StretchEffect types Nader Jawad
Added additional shader stretch effect type to distinguish between shader based stretches that are configured in HWUI as well as those that are configured in both HWUI and SurfaceFlinger Bug: 184297961 Test: manual Change-Id: Ie1083cef7cbaf4f877783d67d429af7f5edd0a3c
2021-05-10Add a linear variant of the stretch effect John Reck
Supports SurfaceViews & everything Test: setprop debug.hwui.stretch_mode 1 Bug: 187718492 Change-Id: I9a222fa4c1a40e80a74cdaf75becb9524cbeed9b
2021-04-10Integrate HWUI with PerformanceHintManager Bo Liu
PerformanceHintManager.Session is in java, so add JNI and a HintSessionWrapper class in HardwareRenderer. Then pass the two calls as two std::functions into DrawFrameTask. Note Session is created per HardwareRenderer, not global (per RenderThread). Session includes UI thread, render thread, and the thread pool. Desired duration is from the intended start duration to the frame deadline. Add an actual frame start time to compute Add system properties: debug.hwui.use_hint_manager to enable PerformanceHintManager debug.hwui.target_cpu_time_percent to control percentage of frame time to be used for target cpu duration. Test: Manual test that there are no crashes and values make sense. Bug: 158791282 Change-Id: I83f25433c10daa20033803fb7c4ae45eab34f1d3
2021-03-17Merge "Use `ro.boot.qemu` to check if the device is an emulator" am: ↵ Treehugger Robot
62d82e0b33 am: 4ba8fde9ba am: de535f6eb2 Original change: https://android-review.googlesource.com/c/platform/frameworks/base/+/1625662 Change-Id: Ia613c23fb779697882f9928b15684fb1ff97e9cd
2021-03-10Use `ro.boot.qemu` to check if the device is an emulator Roman Kiryanov
`ro.kernel.qemu` is deprecated. Bug: 182291166 Test: presubmit Signed-off-by: Roman Kiryanov <rkir@google.com> Change-Id: Id42d79af90b1fe7e182a96c53c84248e81a7c259
2021-02-08Remove render ahead from hwui Ady Abraham
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
2021-02-03Fix render_ahead properties John Reck
0 and -1 both meant default when 0 should mean 0 and -1 should mean default Test: manual Fixes: 179290765 Change-Id: Ia9aa5e3d83757282bfff776e083d6b3d7d29e9c0
2021-01-13Turn on vk unit tests Alec Mouri
* Degrade error when changing the rendering backend from Properties to no longer be fatal. * Support using a single graphics queue for the Vulkan backend. Cloud devices only support a single queue and some Vulkan implementations such as Swiftshader expose only one queue for simplying their implementation, so we'll need to support one. Bug: 175618060 Bug: 162628999 Test: hwui_unit_tests Test: experiment on Pixel 4 enforcing a single queue (settings app, maps, sysui) Change-Id: I495fcabc3c89bd62bbb833998eab4944c6660f6f
generated by cgit v1.2.3-59-g8ed1b (git 2.39.0) at 2025-10-21 18:54:18 +0000