summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2021-09-27[ArrowOS][11.0] base: Add three-fingers-swipe to screenshot [1/2]Henrique Silva
Squashed commit of the following: Author: Henrique Silva <jhenrique09.mcz@hotmail.com> Date: Wed May 8 11:52:38 2019 -0300 base: SwipeToScreenshot: Import MIUI implementation Reverse engineered from Xiaomi/beryllium/beryllium:9/PKQ1.180729.001/9.2.25:user/release-keys Change-Id: I1d5938457b7cc4c5ef1fec235d27d8a40526961a Author: Henrique Silva <jhenrique09.mcz@hotmail.com> Date: Tue May 7 13:29:38 2019 -0300 base: Improvements for swipe to screenshot * Cancel touch events when tap with three fingers * Dyrex2004: Fix typos (Screnshot -> Screenshot) Change-Id: I90c5d73f37f277052c4e4802fd3006f607313068 Author: ghbhaha <ghbhaha@gmail.com> Date: Sun Aug 9 01:23:33 2015 +0800 base: Add three-fingers-swipe to screenshot [1/2] The feature is ported from Oppo ColorOS. With the options on, users can swipe down on the screen to have a screenshot. Original smali port by: wuxianlin Change-Id: I87ed333f81e1259758ea899bf019061d8ab85fd0 Change-Id: I78c67e5470a5024034be92736a36bc43b3ca4487
2021-09-27[crdroid][11.0] base: General workarounds for stitchimagePranav Vashi
Squashed commit of the following: Author: ender1324 <ender1324@abv.bg> Date: Thu, 4 Feb 2021 15:16:33 +0200 Fallback stitchimage's broken value to unknown - For some reason stitchimage returns 9528, no idea why - A proper fix would be to fix it from the apk itself but "we dont do that here" - Thanks to kandnub for helping me find the core issue Fixes: 02-03 22:05:46.956 10753 10753 E AndroidRuntime: java.lang.RuntimeException: Unable to create service com.asus.stitchimage.OverlayService: java.lang.IllegalArgumentException: Invalid usage 9528 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.app.ActivityThread.handleCreateService(ActivityThread.java:4198) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.app.ActivityThread.access$1500(ActivityThread.java:237) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.app.ActivityThread$H.handleMessage(ActivityThread.java:1932) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.os.Handler.dispatchMessage(Handler.java:106) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.os.Looper.loop(Looper.java:223) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.app.ActivityThread.main(ActivityThread.java:7661) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at java.lang.reflect.Method.invoke(Native Method) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run(RuntimeInit.java:592) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:947) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: Caused by: java.lang.IllegalArgumentException: Invalid usage 9528 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.media.AudioAttributes$Builder.setUsage(AudioAttributes.java:783) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.asus.stitchimage.f.g.b(Unknown Source:7) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.asus.stitchimage.f.t.a(Unknown Source:19) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.asus.stitchimage.OverlayService.B(Unknown Source:83) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at com.asus.stitchimage.OverlayService.onCreate(Unknown Source:73) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: at android.app.ActivityThread.handleCreateService(ActivityThread.java:4186) 02-03 22:05:46.956 10753 10753 E AndroidRuntime: ... 8 more Co-authored-by: Kandarp <kandnub@gmail.com> Signed-off-by: Pranav Vashi <neobuddy89@gmail.com> Author: jhenrique09 <jhenrique09.mcz@hotmail.com> Date: Wed, 3 Feb 2021 22:20:43 +0200 base: Whitelist stitchimage Signed-off-by: Pranav Vashi <neobuddy89@gmail.com> Author: Pranav Vashi <neobuddy89@gmail.com> Date: Fri, 14 May 2021 20:22:34 +0530 Fix screenshot sharing with longshot and launcher Signed-off-by: Pranav Vashi <neobuddy89@gmail.com> Author: ender1324 <ender1324@abv.bg> Date: Wed, 3 Feb 2021 15:40:04 +0200 base: Add missing long screenshot permission Signed-off-by: Pranav Vashi <neobuddy89@gmail.com> Signed-off-by: alk3pInjection <webmaster@raspii.tech> Change-Id: I4fd74068c6b634c7db5ae0226a85b11959a532ed
2021-09-27[crdroid][11.0] base: Use Asus Stitchimage as default screenshot serviceLaker
Also add delete action for partial screenshot notification Signed-off-by: Pranav Vashi <neobuddy89@gmail.com> Change-Id: I10880a731b6c91edd32f6a0084ad609e07b2564f
2021-09-27Revert "Implement click to take partial screenshot [1/3]"alk3pInjection
* meaningless since we'll introduce StitchImage This reverts commit 7e0f34b2c43bfa2a8466dccaaf5eb613393aa347. Change-Id: I201f34983e4b2fb77318d1b04ba49036948d1f8b
2021-09-27[crdroid][11.0] Reapply "Post a silent notification if screenshot is dismissed"Pranav Vashi
This reverts commit 5d4cb837cc3cb0be1efa06ed37fc1f55b473c904. neobuddy89: Also remove broken "Delete" action Change-Id: I6d02c88945df5131931e4d337fa8f07e2d1ddba8
2021-09-27[crdroid][11.0] Screenshot: Add delete action chip intentganeshi4u
Signed-off-by: ganeshi4u <ganeshvarmai4u@gmail.com> Change-Id: I4432178f93601af91e2cae5e7206a1764c201aab
2021-09-27[ProtonAOSP][rvc] GlobalScreenshot: Fix old screenshot preview appearing in ↵Danny Lin
screenshots The screenshot code already dismisses the old screenshot preview immediately before taking a new one, but it fails to account for the fact that the change will not take effect until the next frame render, and takes a new screenshot immediately. To fix the problem, we can invalidate the root view just in case to make sure the screen will be re-rendered, and then take the screenshot after the new frame has been committed. Unfortunately, this adds ~2 frames of latency to screenshot capturing because we have to attach the screenshot callback to the beginning of the next frame, but I'm not aware of any way to reduce it as we can't attach a post-frame commit callback from here. Change-Id: I76aaeb27747c5f025e31ccf6dbcff45d2c26d484
2021-09-27[ProtonAOSP][rvc] GlobalScreenshot: Speed up actions expansion animationDanny Lin
400 ms feels too long for this, especially when the actions already take so long to show up due to how long saving the screenshot takes. Change-Id: Ia7a264e34ad417b734716b7f6a57f2f5147d693d
2021-09-27[ProtonAOSP][rvc] GlobalScreenshot: Reduce screenshot dismiss delay to 3 secondsDanny Lin
2 seconds is more than enough for most cases. Keeping the overlay around for too long forces the user to dismiss it manually, which is a regression compared to the old screenshot notification UI. Change-Id: I5644a055d1285d390d77430670fd6f43862c9608 Signed-off-by: Pranav Vashi <neobuddy89@gmail.com>
2021-09-27[ProtonAOSP][rvc] Editor: Disable magnifier position animationDanny Lin
In my opinion, this animation serves no purpose because it doesn't correspond to any animations in the text selection/handle flow, and is barely noticeable in typical text selection cases. In addition, it makes the magnifier feel slow/laggy when selecting a large block of text quickly. Change-Id: I5fed9fc06db4491dc2811815741830001d295030
2021-09-27[ProtonAOSP][rvc] SystemUI: FingerprintDialogView: Use accent color instead ↵jhenrique09
of teal Change-Id: I97ab75e68693c071ec50e78a9159202f83cf284b
2021-09-27[ProtonAOSP][rvc] SystemUI: Don't render snooze undo button in all-capsDanny Lin
We've disabled all-caps button text throughout the system, but this label has been capitalized manually in strings. Make it consistent with the rest of the system. Change-Id: I95d2b97c3b493b4928d0965e82b3272232d5b567
2021-09-27[ProtonAOSP][rvc] SystemUI: Remove divider between QS tiles and media playerDanny Lin
The color used for this divider in dark mode contrasts too much with the pure black background in the QS, and even in light mode, the divider disrupts the otherwise clean and streamlined style of not having dividers in the QS. Let's just remove the divider altogether for both themes to clean up the QS when the media player is shown. The code controlling the divider has been removed altogether to avoid misleading people who read through the code, as well as to reduce unused code bloat. Change-Id: Ie3d9e0070c974b598a8784225375882b4b67173e
2021-09-27[ProtonAOSP][rvc] SystemUI: Define blur radius in device-independent pixelsDanny Lin
Specifying the blur radius in pixels leads to inconsistent results across devices with different pixel densities. With a Pixel 4 XL (1440x3040) as the reference display (3.5x scale factor), 175 px is more or less the amount of blur we want, but it's too strong on the Pixel 5 (1080x2340). Define the blur radius in device-independent pixels (calculated from the reference display described above) to make the strength uniform. Change-Id: I3295282de81389cc863d63d58370febaf16a0623
2021-09-27[ProtonAOSP][rvc] SystemUI: Bump background blur radius to 175 pxDanny Lin
The increased radius helps create a "glass blur" effect, which looks better than the default radius that lies in an ugly intermediate range between "no blur" and "glass blur". Change-Id: I24fc4f0f99468371fa3b16017e458f6151518858
2021-09-27[ProtonAOSP][rvc] LayoutInflater: Opportunistically create views directly ↵Danny Lin
for performance When opening and closing activities in Settings, a significant amount of CPU time is spent performing JNI calls, as reported by simpleperf: 0.39% /system/framework/arm64/boot-framework.oat art_jni_trampoline Reflection in LayoutInflater is responsible for a significant portion of the time spent in the JNI trampoline: 6.08% 0.08% /apex/com.android.art/javalib/arm64/boot.oat art_jni_trampoline | -- art_jni_trampoline | |--12.38%-- java.lang.reflect.Constructor.newInstance | |--0.09%-- [hit in function] | | | |--88.32%-- android.view.LayoutInflater.createView | | | | | |--83.39%-- com.android.internal.policy.PhoneLayoutInflater.onCreateView | | | android.view.LayoutInflater.onCreateView | | | android.view.LayoutInflater.onCreateView | | | android.view.LayoutInflater.createViewFromTag | | | | | | | |--72.73%-- android.view.LayoutInflater.rInflate | | | | | | | | | |--57.90%-- android.view.LayoutInflater.rInflate | | | | | | | | | | | |--94.90%-- android.view.LayoutInflater.inflate | | | | | | android.view.LayoutInflater.inflate | | | | | | |--35.86%-- [hit in function] | | | | | | | | | | | | | |--58.15%-- androidx.preference.PreferenceGroupAdapter.onCreateViewHolder Empirical testing of interacting with ~113 real-world apps reveals that many of the most frequently-inflated views are framework classes: 13486 android.widget.LinearLayout 6930 android.widget.View 6447 android.widget.FrameLayout 5613 android.widget.ViewStub 5608 androidx.constraintlayout.widget.ConstraintLayout 4722 android.widget.TextView 4431 com.google.android.material.textview.MaterialTextView 3570 eu.faircode.email.FixedTextView 3044 android.widget.ImageView 2665 android.widget.RelativeLayout 1694 android.widget.Space 979 androidx.preference.internal.PreferenceImageView 926 androidx.appcompat.view.menu.ActionMenuItemView 884 androidx.appcompat.widget.AppCompatImageView 855 slack.uikit.components.icon.SKIconView 770 android.widget.ProgressBar 743 com.fastaccess.ui.widgets.FontTextView 541 androidx.recyclerview.widget.RecyclerView 442 androidx.appcompat.widget.AppCompatTextView 404 org.mariotaku.twidere.view.MediaPreviewImageView 393 com.moez.QKSMS.common.widget.QkTextView 382 android.widget.Button 365 slack.widgets.core.textview.ClickableLinkTextView 365 slack.uikit.components.avatar.SKAvatarView 352 com.google.android.libraries.inputmethod.widgets.SoftKeyView 351 com.android.launcher3.BubbleTextView 315 slack.widgets.core.viewcontainer.SingleViewContainer 315 slack.widgets.core.textview.MaxWidthTextView 313 androidx.constraintlayout.widget.Barrier 302 slack.app.ui.widgets.ReactionsLayout 302 slack.app.ui.messages.widgets.MessageLayout 302 slack.app.ui.messages.widgets.MessageHeader 290 com.android.launcher3.views.DoubleShadowBubbleTextView 285 com.android.internal.widget.CachingIconView 265 android.widget.ImageButton 262 androidx.constraintlayout.widget.Guideline 249 org.thoughtcrime.securesms.components.emoji.EmojiTextView 234 com.google.android.libraries.inputmethod.widgets.AutoSizeTextView 232 com.android.internal.widget.RemeasuringLinearLayout 228 android.view.ViewStub 227 android.app.ViewStub 226 android.webkit.ViewStub 221 im.vector.app.core.ui.views.ShieldImageView 219 androidx.constraintlayout.widget.Group 214 androidx.coordinatorlayout.widget.CoordinatorLayout 204 androidx.appcompat.widget.ContentFrameLayout All framework classes seen: 13486 android.widget.LinearLayout 6930 android.widget.View 6447 android.widget.FrameLayout 5613 android.widget.ViewStub 4722 android.widget.TextView 3044 android.widget.ImageView 2665 android.widget.RelativeLayout 1694 android.widget.Space 770 android.widget.ProgressBar 382 android.widget.Button 265 android.widget.ImageButton 228 android.view.ViewStub 227 android.app.ViewStub 226 android.webkit.ViewStub 145 android.widget.Switch 117 android.widget.DateTimeView 86 android.widget.Toolbar 68 android.widget.HorizontalScrollView 67 android.widget.ScrollView 65 android.widget.NotificationHeaderView 65 android.webkit.NotificationHeaderView 65 android.view.NotificationHeaderView 65 android.app.NotificationHeaderView 63 android.webkit.View 63 android.view.View 62 android.app.View 58 android.widget.ListView 50 android.widget.QuickContactBadge 40 android.widget.SeekBar 38 android.widget.CheckBox 16 android.widget.GridLayout 15 android.widget.TableRow 15 android.widget.RadioGroup 15 android.widget.Chronometer 13 android.widget.ViewFlipper 9 android.widget.Spinner 8 android.widget.ViewSwitcher 8 android.widget.TextSwitcher 8 android.widget.SurfaceView 8 android.widget.CheckedTextView 8 android.preference.PreferenceFrameLayout 7 android.widget.TwoLineListItem 5 android.widget.TableLayout 5 android.widget.EditText 3 android.widget.TabWidget 3 android.widget.TabHost 2 android.widget.ZoomButton 2 android.widget.TextureView 2 android.widget.ExpandableListView 2 android.webkit.TextureView 2 android.view.TextureView 2 android.app.TextureView 1 android.widget.WebView 1 android.widget.ViewAnimator 1 android.widget.TextClock 1 android.widget.AutoCompleteTextView 1 android.webkit.WebView 1 android.webkit.SurfaceView 1 android.view.SurfaceView 1 android.app.SurfaceView Unfortunately, replacing reflection with MethodHandle constructors is counter-productive in terms of performance: Constructor direct: create=5 invoke=42 Constructor reflection: create=310 invoke=433 Constructor MethodHandle: create=3283 invoke=3489 Constructor MethodHandle-exact: create=3273 invoke=3453 To reduce the performance impact of slow reflection, we can leverage the fact that the most frequently-inflated classes are from the framework, and hard-code direct constructor references for them in a switch-case block. Reflection will automatically be used as a fallback for custom app views. Test: simpleperf record -a; verify that Constructor.newInstance -> LayoutInflater no longer appears at the top under art_jni_trampoline Change-Id: I8fcc0e05813ff9ecf1eddca3cc6920e747adf4fc
2021-09-27[ProtonAOSP][rvc] SystemServiceRegistry: Replace ArrayMap with HashMap for ↵Danny Lin
performance When opening and closing activities in Settings, a significant amount of CPU time is spent looking up ArrayMap entries, as reported by simpleperf: 0.12% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes PackageManagerService is responsible for a significant portion of the time spent in ArrayMap lookups: 0.08% 0.08% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes | -- android.util.ArrayMap.binarySearchHashes | --50.00%-- android.util.ArrayMap.indexOf | |--36.71%-- android.util.ArrayMap.get | |--0.87%-- [hit in function] | | | |--5.42%-- android.app.SystemServiceRegistry.getSystemService | | android.app.ContextImpl.getSystemService | | android.view.ContextThemeWrapper.getSystemService | | android.app.Activity.getSystemService | | | | | |--52.18%-- TemporaryFile-kEdnnv[+9b97baa8] | | | TemporaryFile-FwF2he[+9b96d048] | | | art_quick_invoke_stub | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | art::interpreter::ArtInterpreterToCompiledCodeBridge(art::Thread*, art::ArtMethod*, art::ShadowFrame*, unsigned short, art::JValue*) | | | bool art::interpreter::DoCall<false, false>(art::ArtMethod*, art::Thread*, art::ShadowFrame&, art::Instruction const*, unsigned short, art::JValue*) | | | MterpInvokeVirtual | | | mterp_op_invoke_virtual | | | libcrypto.so[+3fac6] | | | MterpInvokeDirect | | | mterp_op_invoke_direct | | | libcrypto.so[+3faa8] | | | MterpInvokeVirtual | | | mterp_op_invoke_virtual | | | libcrypto.so[+3730c] | | | art::interpreter::Execute(art::Thread*, art::CodeItemDataAccessor const&, art::ShadowFrame&, art::JValue, bool, bool) (.llvm.13341777805210357670) | | | artQuickToInterpreterBridge | | | art_quick_to_interpreter_bridge | | | java.util.concurrent.Executors$RunnableAdapter.call | | | java.util.concurrent.FutureTask.run | | | java.util.concurrent.ThreadPoolExecutor.runWorker | | | java.util.concurrent.ThreadPoolExecutor$Worker.run | | | java.lang.Thread.run | | | art_quick_invoke_stub | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | art::JValue art::InvokeVirtualOrInterfaceWithJValues<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, jvalue const*) | | | art::Thread::CreateCallback(void*) | | | __pthread_start(void*) | | | __start_thread | | | | | --47.82%-- android.view.ContextThemeWrapper.getSystemService | | android.view.ContextThemeWrapper.getSystemService | | android.content.Context.getSystemService | | android.view.View.onVisibilityAggregated | | android.view.View.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewGroup.dispatchAttachedToWindow | | android.view.ViewRootImpl.performTraversals | | android.view.ViewRootImpl.doTraversal | | android.content.ContextWrapper.getAssets [DEDUPED] | | android.view.Choreographer.doCallbacks | | android.view.Choreographer.doFrame | | android.view.Choreographer$FrameDisplayEventReceiver.run | | android.os.Handler.dispatchMessage | | android.os.Looper.loop | | android.app.ActivityThread.main | | art_quick_invoke_static_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::InvokeMethod(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jobject*, _jobject*, unsigned long) | | art::Method_invoke(_JNIEnv*, _jobject*, _jobject*, _jobjectArray*) | | art_jni_trampoline | | com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run | | com.android.internal.os.ZygoteInit.main | | art_quick_invoke_static_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::JValue art::InvokeWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | art::JValue art::InvokeWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | art::JNI<true>::CallStaticVoidMethodV(_JNIEnv*, _jclass*, _jmethodID*, std::__va_list) | | _JNIEnv::CallStaticVoidMethod(_jclass*, _jmethodID*, ...) | | android::AndroidRuntime::start(char const*, android::Vector<android::String8> const&, bool) | | main | | __libc_init | | Empirical testing reveals that SYSTEM_SERVICE_FETCHERS contains 134 entries, at which HashMap is 54% faster than ArrayMap for lookups and 74% faster [1] for insertions. The increased memory usage should be a worthwhile trade-off at this size, so we can safely convert the map to a HashMap in order to improve performance in this hotpath. Because SYSTEM_SERVICE_NAMES, SYSTEM_SERVICE_FETCHERS, and SYSTEM_SERVICE_CLASS_NAMES have the same names and similar uses, all of them have been converted to HashMaps for consistency and performance. [1] https://docs.google.com/spreadsheets/d/136UJS2yVlZyPx30KDNgj4AWldkp9xbzIcWkLFj9RGgk/edit Test: simpleperf record -a; verify that SystemServiceRegistry no longer appears under ArrayMap.binarySearchHashes Change-Id: I2a5b23793a4fc8aa720eead3ecc7ca4589cb67da
2021-09-27[ProtonAOSP][rvc] PackageManagerService: Replace ArrayMap with HashMap for ↵Danny Lin
performance When opening and closing activities in Settings, a significant amount of CPU time is spent looking up ArrayMap entries, as reported by simpleperf: 0.12% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes PackageManagerService is responsible for a significant portion of the time spent in ArrayMap lookups: 0.08% 0.08% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes | -- android.util.ArrayMap.binarySearchHashes | --50.00%-- android.util.ArrayMap.indexOf | |--36.71%-- android.util.ArrayMap.get | |--0.87%-- [hit in function] | | | |--8.21%-- com.android.server.pm.PackageManagerService$PackageManagerInternalImpl.getPackage | | | | | |--74.06%-- com.android.server.pm.ComponentResolver$ActivityIntentResolver.newResult | | | | | --25.94%-- com.android.server.usage.AppStandbyController$Injector.getValidCrossProfileTargets | | | |--7.13%-- com.android.server.pm.PackageManagerService.resolveInternalPackageNameInternalLocked | | | | | |--56.77%-- com.android.server.pm.PackageManagerService.getPackageInfoInternal | | | | | | | |--54.96%-- com.android.server.power.PowerManagerService.acquireWakeLockInternal | | | | | | | --45.04%-- com.android.server.power.PowerManagerService.userActivityInternal | | | | | |--22.44%-- com.android.server.pm.PackageManagerService$PackageManagerInternalImpl.getPackage | | | | | --20.79%-- com.android.server.pm.PackageManagerService.getPackageSettingInternal | | | |--7.08%-- com.android.server.pm.PackageManagerService.getPackageSettingInternal | | | | | |--40.25%-- com.android.server.pm.PackageManagerService.getPackageUidInternal | | | | | |--33.33%-- com.android.server.pm.PackageManagerService.getPackageSetting | | | | | | | |--63.07%-- com.android.server.pm.PackageManagerService$PackageManagerInternalImpl.filterAppAccess | | | | | | | --36.93%-- com.android.server.pm.permission.PermissionManagerService.checkPermissionInternal | | | | | --26.42%-- com.android.server.pm.PackageManagerService.applyPostResolutionFilter Empirical testing reveals that mPackages contains 359 entries on my device, at which HashMap is 64% faster than ArrayMap for lookups and 85% faster [1] for insertions. The increased memory usage should be a worthwhile trade-off at this size, so we can safely convert the map to a HashMap in order to improve performance in this hotpath. Because profiling sample hits can also be seen in Settings.mPackages, which is the same size as PackageManagerService.mPackages, both have been converted to HashMaps. [1] https://docs.google.com/spreadsheets/d/136UJS2yVlZyPx30KDNgj4AWldkp9xbzIcWkLFj9RGgk/edit Test: simpleperf record -a; verify that PackageManagerService no longer appears under ArrayMap.binarySearchHashes Change-Id: Ie888a6d88390529fde7cb4ba4758058c74b1ff01
2021-09-27[ProtonAOSP][rvc] InsetsStateController: Replace ArrayMap with HashMap for ↵Danny Lin
performance When opening and closing activities in Settings, a significant amount of CPU time is spent looking up ArrayMap entries, as reported by simpleperf: 0.12% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes InsetsStateController is responsible for a significant portion of the time spent in ArrayMap lookups: 0.08% 0.08% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes | -- android.util.ArrayMap.binarySearchHashes | --50.00%-- android.util.ArrayMap.indexOf | |--36.71%-- android.util.ArrayMap.get | |--0.87%-- [hit in function] | | | |--6.67%-- com.android.server.wm.InsetsStateController.peekSourceProvider | | | | | |--68.77%-- com.android.server.wm.DisplayPolicy.beginLayoutLw | | | com.android.server.wm.DisplayContent.performLayoutNoTrace | | | com.android.server.wm.DisplayContent.performLayout | | | com.android.server.wm.DisplayContent.applySurfaceChangesTransaction | | | com.android.server.wm.RootWindowContainer.applySurfaceChangesTransaction | | | com.android.server.wm.RootWindowContainer.performSurfacePlacementNoTrace | | | com.android.server.wm.RootWindowContainer.performSurfacePlacement | | | com.android.server.wm.WindowSurfacePlacer.performSurfacePlacementLoop | | | com.android.server.wm.WindowSurfacePlacer.performSurfacePlacement | | | com.android.server.wm.WindowManagerService.relayoutWindow | | | com.android.server.wm.Session.relayout | | | android.view.IWindowSession$Stub.onTransact | | | com.android.server.wm.Session.onTransact | | | android.os.Binder.execTransactInternal | | | android.os.Binder.execTransact | | | art_quick_invoke_stub | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | | art::JNI<false>::CallBooleanMethodV(_JNIEnv*, _jobject*, _jmethodID*, std::__va_list) | | | _JNIEnv::CallBooleanMethod(_jobject*, _jmethodID*, ...) | | | JavaBBinder::onTransact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | | android::BBinder::transact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | | android::IPCThreadState::executeCommand(int) | | | android::IPCThreadState::getAndExecuteCommand() | | | android::IPCThreadState::joinThreadPool(bool) | | | android::PoolThread::threadLoop() | | | android::Thread::_threadLoop(void*) | | | android::AndroidRuntime::javaThreadShell(void*) | | | thread_data_t::trampoline(thread_data_t const*) | | | __pthread_start(void*) | | | __start_thread | | | | | --31.23%-- com.android.server.wm.DisplayPolicy.updateHideNavInputEventReceiver | | com.android.server.wm.InsetsPolicy.updateBarControlTarget | | com.android.server.wm.DisplayPolicy.updateSystemUiVisibilityLw | | com.android.server.wm.DisplayPolicy.finishPostLayoutPolicyLw | | com.android.server.wm.DisplayContent.applySurfaceChangesTransaction | | com.android.server.wm.RootWindowContainer.applySurfaceChangesTransaction | | com.android.server.wm.RootWindowContainer.performSurfacePlacementNoTrace | | com.android.server.wm.RootWindowContainer.performSurfacePlacement | | com.android.server.wm.WindowSurfacePlacer.performSurfacePlacementLoop | | com.android.server.wm.WindowSurfacePlacer.performSurfacePlacement | | com.android.server.wm.WindowManagerService.postWindowRemoveCleanupLocked | | com.android.server.wm.WindowState.removeImmediately | | com.android.server.wm.WindowState.removeIfPossible | | com.android.server.wm.WindowState.removeIfPossible | | com.android.server.wm.WindowManagerService.removeWindow | | com.android.server.wm.Session.remove | | android.view.IWindowSession$Stub.onTransact | | com.android.server.wm.Session.onTransact | | android.os.Binder.execTransactInternal | | android.os.Binder.execTransact | | art_quick_invoke_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | art::JNI<false>::CallBooleanMethodV(_JNIEnv*, _jobject*, _jmethodID*, std::__va_list) | | _JNIEnv::CallBooleanMethod(_jobject*, _jmethodID*, ...) | | JavaBBinder::onTransact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::BBinder::transact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::IPCThreadState::executeCommand(int) | | android::IPCThreadState::getAndExecuteCommand() | | android::IPCThreadState::joinThreadPool(bool) | | android::PoolThread::threadLoop() | | android::Thread::_threadLoop(void*) | | android::AndroidRuntime::javaThreadShell(void*) | | thread_data_t::trampoline(thread_data_t const*) | | __pthread_start(void*) | | __start_thread | | Empirical testing reveals that mProviders usually contains 9 entries, at which HashMap is 34% faster than ArrayMap for lookups and 57% faster [1] for insertions. The increased memory usage should be negligible at this size, so we can safely convert the map to a HashMap in order to improve performance in this hotpath. [1] https://docs.google.com/spreadsheets/d/136UJS2yVlZyPx30KDNgj4AWldkp9xbzIcWkLFj9RGgk/edit Test: simpleperf record -a; verify that InsetsStateController no longer appears under ArrayMap.binarySearchHashes Change-Id: Ic08d4c3e56cf10b322eabc115de441c6c5f4a898
2021-09-27[ProtonAOSP][rvc] LocalServices: Replace ArrayMap with HashMap for performanceDanny Lin
When opening and closing activities in Settings, a significant amount of CPU time is spent looking up ArrayMap entries, as reported by simpleperf: 0.12% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes LocalServices is responsible for a significant portion of the time spent in ArrayMap lookups: 0.08% 0.08% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes | -- android.util.ArrayMap.binarySearchHashes | --50.00%-- android.util.ArrayMap.indexOf | |--36.71%-- android.util.ArrayMap.get | |--0.87%-- [hit in function] | | | |--6.12%-- com.android.server.LocalServices.getService | | | | | |--42.41%-- com.android.server.inputmethod.InputMethodManagerInternal.get [DEDUPED] | | | com.android.server.wm.-$$Lambda$DisplayContent$-xtu90EUfC_AM8Qe5g8vDDI07_E.run | | | android.os.Handler.dispatchMessage | | | android.os.Looper.loop | | | android.os.HandlerThread.run | | | com.android.server.ServiceThread.run | | | art_quick_invoke_stub | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | art::JValue art::InvokeVirtualOrInterfaceWithJValues<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, jvalue const*) | | | art::Thread::CreateCallback(void*) | | | __pthread_start(void*) | | | __start_thread | | | | | |--36.66%-- com.android.server.notification.NotificationManagerService.enqueueNotificationInternal | | | com.android.server.notification.NotificationManagerService.enqueueNotificationInternal | | | com.android.server.notification.NotificationManagerService$10.enqueueNotificationWithTag | | | android.app.INotificationManager$Stub.onTransact | | | android.os.Binder.execTransactInternal | | | android.os.Binder.execTransact | | | art_quick_invoke_stub | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | | art::JNI<false>::CallBooleanMethodV(_JNIEnv*, _jobject*, _jmethodID*, std::__va_list) | | | _JNIEnv::CallBooleanMethod(_jobject*, _jmethodID*, ...) | | | JavaBBinder::onTransact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | | android::BBinder::transact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | | android::IPCThreadState::executeCommand(int) | | | android::IPCThreadState::getAndExecuteCommand() | | | android::IPCThreadState::joinThreadPool(bool) | | | android::PoolThread::threadLoop() | | | android::Thread::_threadLoop(void*) | | | android::AndroidRuntime::javaThreadShell(void*) | | | thread_data_t::trampoline(thread_data_t const*) | | | __pthread_start(void*) | | | __start_thread | | | | | --20.93%-- com.android.server.oemlock.OemLockService.setPersistentDataBlockOemUnlockAllowedBit | | com.android.server.oemlock.OemLockService$2.isOemUnlockAllowed | | android.service.oemlock.IOemLockService$Stub.onTransact | | android.os.Binder.execTransactInternal | | android.os.Binder.execTransact | | art_quick_invoke_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | art::JNI<false>::CallBooleanMethodV(_JNIEnv*, _jobject*, _jmethodID*, std::__va_list) | | _JNIEnv::CallBooleanMethod(_jobject*, _jmethodID*, ...) | | JavaBBinder::onTransact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::BBinder::transact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::IPCThreadState::executeCommand(int) | | android::IPCThreadState::getAndExecuteCommand() | | android::IPCThreadState::joinThreadPool(bool) | | android::PoolThread::threadLoop() | | android::Thread::_threadLoop(void*) | | android::AndroidRuntime::javaThreadShell(void*) | | thread_data_t::trampoline(thread_data_t const*) | | __pthread_start(void*) | | __start_thread Empirical testing reveals that sLocalServiceObjects usually contains 68 entries, at which HashMap is 47% faster than ArrayMap for lookups and 68% faster [1] for insertions. The increased memory usage should be negligible at this size, so we can safely convert the map to a HashMap in order to improve performance in this hotpath. [1] https://docs.google.com/spreadsheets/d/136UJS2yVlZyPx30KDNgj4AWldkp9xbzIcWkLFj9RGgk/edit Test: simpleperf record -a; verify that LocalServices no longer appears under ArrayMap.binarySearchHashes Change-Id: Ifd1f8b7940eba7723f93a73456470a84d34656ae
2021-09-27[ProtonAOSP][rvc] ThemedResourceCache: Replace ArrayMap with HashMap for ↵Danny Lin
performance When opening and closing activities in Settings, a significant amount of CPU time is spent looking up ArrayMap entries, as reported by simpleperf: 0.12% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes ThemedResourceCache is responsible for a significant portion of the time spent in ArrayMap lookups: 0.08% 0.08% /system/framework/arm64/boot-framework.oat android.util.ArrayMap.binarySearchHashes | -- android.util.ArrayMap.binarySearchHashes | --50.00%-- android.util.ArrayMap.indexOf | |--36.71%-- android.util.ArrayMap.get | |--0.87%-- [hit in function] | | | |--9.64%-- android.content.res.ThemedResourceCache.getThemedLocked | | android.content.res.ThemedResourceCache.get | | | | | |--77.92%-- android.content.res.DrawableCache.getInstance | | | android.content.res.ResourcesImpl.loadDrawable | | | android.content.res.Resources.loadDrawable | | | android.content.res.TypedArray.getDrawableForDensity | | | android.content.res.Resources.getColor [DEDUPED] | | | | | | | |--62.94%-- android.view.View.<init> | | | | | | | | | |--64.58%-- android.view.ViewGroup.<init> | | | | | android.widget.LinearLayout.<init> | | | | | android.widget.LinearLayout.<init> | | | | | art_quick_invoke_stub | | | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | | | art::InvokeConstructor(art::ScopedObjectAccessAlreadyRunnable const&, art::ArtMethod*, art::ObjPtr<art::mirror::Object>, _jobject*) | | | | | art::Constructor_newInstance0(_JNIEnv*, _jobject*, _jobjectArray*) | | | | | art_jni_trampoline | | | | | java.lang.reflect.Constructor.newInstance | | | | | android.view.LayoutInflater.createView | | | | | com.android.internal.policy.PhoneLayoutInflater.onCreateView | | | | | android.view.LayoutInflater.onCreateView | | | | | android.view.LayoutInflater.onCreateView | | | | | android.view.LayoutInflater.createViewFromTag | | | | | android.view.LayoutInflater.inflate | | | | | android.view.LayoutInflater.inflate | | | | | | | | | --35.42%-- android.widget.TextView.<init> | | | | android.widget.Button.<init> | | | | art_quick_invoke_stub | | | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | | | art::InvokeConstructor(art::ScopedObjectAccessAlreadyRunnable const&, art::ArtMethod*, art::ObjPtr<art::mirror::Object>, _jobject*) | | | | art::Constructor_newInstance0(_JNIEnv*, _jobject*, _jobjectArray*) | | | | art_jni_trampoline | | | | java.lang.reflect.Constructor.newInstance | | | | android.view.LayoutInflater.createView | | | | com.android.internal.policy.PhoneLayoutInflater.onCreateView | | | | android.view.LayoutInflater.onCreateView | | | | android.view.LayoutInflater.onCreateView | | | | android.view.LayoutInflater.createViewFromTag | | | | android.view.LayoutInflater.rInflate | | | | android.view.LayoutInflater.rInflate | | | | android.view.LayoutInflater.rInflate | | | | android.view.LayoutInflater.inflate | | | | android.view.LayoutInflater.inflate | | | | android.view.LayoutInflater.inflate | | | | | | | --37.06%-- com.android.internal.widget.ToolbarWidgetWrapper.<init> | | | | | --22.08%-- android.content.res.ConfigurationBoundResourceCache.get | | android.content.res.ConfigurationBoundResourceCache.getInstance | | android.content.res.ResourcesImpl.loadComplexColorFromName | | android.content.res.ResourcesImpl.loadColorStateList | | android.content.res.Resources.loadColorStateList | | android.content.res.TypedArray.getColorStateList | | android.widget.TextView.readTextAppearance | | android.widget.TextView.setTextAppearance | | android.widget.TextView.setTextAppearance | | android.widget.Toolbar.setTitle | | com.android.wifi.x.com.android.internal.util.StateMachine$SmHandler.handleMessage | | android.view.SurfaceControl.copyFrom Empirical testing reveals that mThemedEntries usually contains around 14 entries, at which HashMap is 35% faster than ArrayMap for lookups and 54% faster [1] for insertions. The increased memory usage should be negligible at this size, so we can safely convert the map to a HashMap in order to improve performance in this hotpath. [1] https://docs.google.com/spreadsheets/d/136UJS2yVlZyPx30KDNgj4AWldkp9xbzIcWkLFj9RGgk/edit Test: simpleperf record -a; verify that ThemedResourceCache no longer appears under ArrayMap.binarySearchHashes Change-Id: I39e1c4b03fe0e60f933f02e253d2d3c4a483146f
2021-09-27[ProtonAOSP][rvc] Trace: Disable debug tracing in production buildsDanny Lin
When opening and closing activities in Settings, a significant amount of CPU time is spent checking whether ATrace tags are enabled, as measured by simpleperf: 0.12% /system/lib64/libcutils.so atrace_get_enabled_tag android.os.Trace is responsible for a significant portion of the time spent in the checks: 0.10% 0.07% /system/lib64/libcutils.so atrace_get_enabled_tags | -- atrace_get_enabled_tags | | |--1.62%-- android.os.Trace.traceEnd | |--36.90%-- [hit in function] | | | |--29.76%-- android.view.Choreographer.doCallbacks | | android.view.Choreographer.doFrame | | android.view.Choreographer$FrameDisplayEventReceiver.run | | android.os.Handler.dispatchMessage | | android.os.Looper.loop | | android.os.HandlerThread.run | | com.android.server.ServiceThread.run | | art_quick_invoke_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::JValue art::InvokeVirtualOrInterfaceWithJValues<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, jvalue const*) | | art::Thread::CreateCallback(void*) | | __pthread_start(void*) | | __start_thread | | | |--19.00%-- com.android.server.BatteryService$BatteryPropertiesRegistrar.getProperty | | android.os.IBatteryPropertiesRegistrar$Stub.onTransact | | android.os.Binder.execTransactInternal | | android.os.Binder.execTransact | | art_quick_invoke_stub | | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | | art::JValue art::InvokeVirtualOrInterfaceWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | | art::JNI<false>::CallBooleanMethodV(_JNIEnv*, _jobject*, _jmethodID*, std::__va_list) | | _JNIEnv::CallBooleanMethod(_jobject*, _jmethodID*, ...) | | JavaBBinder::onTransact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::BBinder::transact(unsigned int, android::Parcel const&, android::Parcel*, unsigned int) | | android::IPCThreadState::executeCommand(int) | | android::IPCThreadState::getAndExecuteCommand() | | android::IPCThreadState::joinThreadPool(bool) | | android::PoolThread::threadLoop() | | android::Thread::_threadLoop(void*) | | android::AndroidRuntime::javaThreadShell(void*) | | thread_data_t::trampoline(thread_data_t const*) | | __pthread_start(void*) | | __start_thread | | | --14.33%-- android.view.ViewRootImpl.performDraw | android.view.ViewRootImpl.performTraversals | android.view.ViewRootImpl.doTraversal | android.content.ContextWrapper.getAssets [DEDUPED] | android.view.Choreographer.doCallbacks | android.view.Choreographer.doFrame | android.view.Choreographer$FrameDisplayEventReceiver.run | android.os.Handler.dispatchMessage | android.os.Looper.loop | android.app.ActivityThread.main | art_quick_invoke_static_stub | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | art::InvokeMethod(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jobject*, _jobject*, unsigned long) | art::Method_invoke(_JNIEnv*, _jobject*, _jobject*, _jobjectArray*) | art_jni_trampoline | com.android.internal.os.RuntimeInit$MethodAndArgsCaller.run | com.android.internal.os.ZygoteInit.main | art_quick_invoke_static_stub | art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) | art::JValue art::InvokeWithVarArgs<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, std::__va_list) | art::JValue art::InvokeWithVarArgs<_jmethodID*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, _jmethodID*, std::__va_list) | art::JNI<true>::CallStaticVoidMethodV(_JNIEnv*, _jclass*, _jmethodID*, std::__va_list) | _JNIEnv::CallStaticVoidMethod(_jclass*, _jmethodID*, ...) | android::AndroidRuntime::start(char const*, android::Vector<android::String8> const&, bool) | main | __libc_init We're unlikely to trace the system and framework in production systems, so disable debug tracing in non-debuggable builds. The ro.debuggable check is performed from the Java side for simplicity. Test: simpleperf record -a; verify that android.os.Trace no longer appears under atrace_get_enabled_tags Change-Id: I9b63d386c041ffd9d7ff34ecd0ec1eb95dc03b4b
2021-09-27[ProtonAOSP][rvc] PackageInfo: Optimize ApplicationInfo creationDanny Lin
When opening and closing activities in Settings, a significant amount of CPU time is spent in ART interface method call trampolines, as reported by simpleperf: 0.32% /apex/com.android.art/lib64/libart.so art_quick_imt_conflict_trampoline PackageInfoWithoutStateUtils is responsible for a substantial portion of the time: 0.34% 0.24% /apex/com.android.art/lib64/libart.so art_quick_imt_conflict_trampoline | -- art_quick_imt_conflict_trampoline | |--5.48%-- android.content.pm.parsing.PackageInfoWithoutStateUtils.appInfoFlags | com.android.server.pm.parsing.pkg.PackageImpl.toAppInfoWithoutState | android.content.pm.parsing.PackageInfoWithoutStateUtils.generateApplicationInfoUnchecked | com.android.server.pm.parsing.PackageInfoUtils.generateApplicationInfo | | | |--33.53%-- com.android.server.pm.parsing.PackageInfoUtils.generateActivityInfo | | | | | |--76.24%-- com.android.server.pm.ComponentResolver$ActivityIntentResolver.newResult To avoid the overhead of calling methods through interfaces, opportunistically cast ParsingPackageRead objects to the real implementation and access fields directly on it. This isn't pretty, but it reduces the CPU time wasted on interface method calls. Test: simpleperf record -a; verify that PackageInfoWithoutStateUtils.appInfoFlags no longer appears under art_quick_imt_conflict_trampoline Change-Id: I475ba804c61739c7537e664b09973665f001270b
2021-09-27[ProtonAOSP][rvc] Revert "Pre-emptively take a snapshot when finishing an ↵Danny Lin
activity before changing visibility" This reverts commit 6dad90e5883db82c345a3ab592b26f0fd69fe28d. When opening and closing activities in Settings, a significant amount of CPU time is spent rendering and compressing JPEG screenshots, as reported by simpleperf: 0.46% /system/lib64/libjpeg.so encode_mcu_gather 0.37% /system/lib64/libhwui.so neon::S32_alpha_D32_filter_DX(SkBitmapProcState const&, unsigned int const*, int, unsigned int*) 0.29% /system/lib64/libjpeg.so jsimd_extrgbx_ycc_convert_neon 0.27% /system/lib64/libjpeg.so jsimd_fdct_islow_neon 0.23% /system/lib64/libjpeg.so jsimd_huff_encode_one_block_neon 0.14% /system/lib64/libjpeg.so jsimd_quantize_neon Call graph tracing reveals that TaskSnapshotPersister is responsible for taking the screenshots: 0.16% 0.16% /system/lib64/libjpeg.so encode_mcu_gather | -- encode_mcu_gather | --50.00%-- compress_output process_data_simple_main jpeg_write_scanlines SkJpegEncoder::onEncodeRows(int) SkJpegEncoder::Encode(SkWStream*, SkPixmap const&, SkJpegEncoder::Options const&) SkEncodeImage(SkWStream*, SkPixmap const&, SkEncodedImageFormat, int) android::Bitmap::compress(SkBitmap const&, android::Bitmap::JavaCompressFormat, int, SkWStream*) android::Bitmap::compress(android::Bitmap::JavaCompressFormat, int, SkWStream*) Bitmap_compress(_JNIEnv*, _jobject*, long, int, int, _jobject*, _jbyteArray*) art_jni_trampoline android.graphics.Bitmap.compress com.android.server.wm.TaskSnapshotPersister$StoreWriteQueueItem.writeBuffer com.android.server.wm.TaskSnapshotPersister$StoreWriteQueueItem.write com.android.server.wm.TaskSnapshotPersister$1.run art_quick_invoke_stub art::ArtMethod::Invoke(art::Thread*, unsigned int*, unsigned int, art::JValue*, char const*) art::JValue art::InvokeVirtualOrInterfaceWithJValues<art::ArtMethod*>(art::ScopedObjectAccessAlreadyRunnable const&, _jobject*, art::ArtMethod*, jvalue const*) art::Thread::CreateCallback(void*) __pthread_start(void*) __start_thread Manual code tracing leads to ActivityRecord as the culprit, as it takes a screenshot every time an Activity is finished. This doesn't appear to be critical, so revert the commit that added it in order to reduce excess CPU usage. Test: simpleperf record -a; verify that libjpeg-turbo no longer appears in top sample hits Change-Id: Ib161679f1f3b83787f90b8ef7dcf46d246bb7c57
2021-09-27[ProtonAOSP][rvc] KeyStore: Block key attestation for Google Play ServicesDanny Lin
In order to enforce SafetyNet security, Google Play Services is now using hardware attestation for ctsProfile validation in all cases, even when basic attestation is selected. The SafetyNet API response from GMS will report that basic attestation was used, but under the hood, hardware attestation is always used regardless of the reported state. This results in SafetyNet failing to pass due to TrustZone reporting an unlocked bootloader (and a partially invalidated root of trust) in the key attestation result. We can still take advantage of the fact that this usage of hardware attestation is opportunistic - that is, it falls back to basic attestation if key attestation fails to run - and prevent GMS from using key attestation at the framework level. This causes it to gracefully fall back to basic attestation and pass SafetyNet with an unlocked bootloader. Key attestation is still available for other apps, as there are valid uses for it that do not involve SafetyNet. The "not implemented" error code from keymaster is used to simulate the most realistic failure condition to evade detection, i.e. an old device that lacks support for key attestation. Change-Id: I7282ab22b933434bb11037743d46b8a20dad063a
2021-09-27[master] Fix issue: https://issuetracker.google.com/issues/183024999Tang Ding
There's a performance issue during continually scroll listview when listview in Fling state . Consider such scene as blew: 1.Listview holds a lots of data 2.mTouchMode in TOUCH_MODE_FLING state. 3.Touch Event delivery down event to listview , mTouchMode change to TOUCH_MODE_SCROLL , post checkFlyWheel(40ms timeout). move event make checkFlywheel.endFling() delay to execute , before endFling() satisfied to execute up event occured, flingrunnable.start() will post a extra FlingRunnable in onTouchUp(). Finally Choreographer.CallbackQueue will get more and more FlingRunnable animation callback when keep repeating above steps#3, Choreographer#doFrame#animation will execute FlingRunnbale N times per frame , it will greater than vsync period cause jank (N means repeat steps#3 times). Solution: Remove unnecessary FlingRunnable before add FlingRunnable. Change-Id: I2cbab9171ffbaad87a8908eef09f813cf89ad85a
2021-09-27[master] Remove malloc/free for inline overlay valuesRyan Mitchell
Remove malloc/free of android::ResTable_entry for inline overlay values. Add `target_entry_inline` to the idmap format to encode inline overlay values separate from direct mapping of target resource to overlay resource. This reduces the number of bytes needed to represent a direct mapping of target resource to overlay resource from 9 bytes to 8 bytes per entry. Fixed all idmap alignment issues that required the framework to use "#pragma pack(push, 1)" when loading idmaps. Bug: 170341022 Test: idmap2_tests and libandroidfw_tests Change-Id: Iab4d3902508f02773464724913e0ee966e3689e4
2021-09-27[master] API and boot image updates to import RI BigIntegerHans Boehm
This reorders methods, omits a redundant "implements" spec, and adds <x>ValueExact() methods, which are in OpenJDK 8, but hadn't been implemented on Android. Also update the boot image profile to compile BigInteger-required methods that used to be native. Some of these must be compiled to avoid ART test timeouts. Bug: 136887041 Bug: 119491938 Bug: 28214673 Bug: 28251030 Bug: 2950143 Test: AOSP Boots. Ran some manual Calculator tests on Cuttlefish. Change-Id: Idd72ec6a6fd54ed4ad96a82b26840b41a3882678
2021-09-27[master] Add dalvik.vm.force-java-zygote-fork-loopHans Boehm
Bug: 192020504 Test: Check logcat with and without the flag. Change-Id: Ia46a011125a6e245bbd9065d708d1f905d56e3af
2021-09-27[master] Allow app zygote preload to retain files across forkEgor Pasko
The bug proposes to 'move' the /proc/self/fd/ readlink/stat/etc checks performed by the FileDescriptorAllowlist from before-fork to an earlier stage. The original aim was to allow the app zygote Preload hook to open ashmem/memfd read-only regions to save more RAM (around 5MiB on aarch64) via sharing more across processes. Potentially other files/sockets can be opened - the app zygote takes responsibility of managing file descriptor access controls across its own processes. App Zygote Preload does not run 3rd party code. Unfortunately a straightforward move of the checks to just-before-preload has disadvantages: * opens more codepaths for potential accidental misuse (the zygote accepts commands between preload and fork, there are valid usecases for extending these commands) * this way FileDescriptorAllowlist would need to support more file descriptor types (sockets and maybe pipes), which is not needed now because these FDs are closed right before forking The solution proposed here is to: 1. Determine the set of file descriptors open before preload 2. Run the preload hook 3. Determine FDs opened by the hook and allow them to remain open across fork 4. Hypothetical new attempts to preload (if ever supported) will not affect the allowed FDs - the preload will be able to toss its own FDs the way it wants, but not open the new-new ones Bug: 184808875 Test: Manual: unreleased Chrome patch: while in app zygote preload, create ashmem region, passes it to 'untrusted_app' (=browser process), and call mmap(2) on it. Change-Id: Ie302eabca83a0e4f409cb131e4308b73e5f6a580
2021-09-27[master] Fix error formatting issuesHans Boehm
I missed a couple of CREATE_ERROR calls the first time around. Unfortunately, type checking is loose enough to miss that. Bug: 187992348 Test: TreeHugger Change-Id: Ib7c227fa0e8491d8ce7cd0d82217098664e3182d
2021-09-27[master] Don't fork USAPs with open argument bufferHans Boehm
Delay execution of any zygote command that may end up forking USAPs until we've released our own ZygoteCommandBuffer. Otherwise we fork the child with an open ZygoteCommandBuffer, preventing it from opening its own. Bug: 182084594 Test: Build and boot AOSP, with USAPs enabled and disabled. Change-Id: I2366367d2155a120092fbdcae13cbb3d7f4e1d16
2021-09-27[master] Use tmpArgBuffer correctly when argBuffer is nullWang Han
* Seems to be a typo and can cause NPE. Change-Id: I7e253c97e02767a076d9450e88ccea08c5122bb1
2021-09-27[master] More consistently retry system calls on EINTRHans Boehm
We were not retrying an accept() call om EINTR, resulting in occasional zygote failures. This fixes that, and a few other calls documented to potentially return EINTR. This is based on a quick seacrh of the two files affected by this CL. Bug: 193753947 Bug: 187992348 Test: Build and boot AOSP Change-Id: Icbfb38be5110607c121545e5c200ce65d1eefbfe
2021-09-27[master] Add zygote native fork loopHans Boehm
Do not return to Java mode between consecutive fork operations. This greatly reduces the Zygote overhead, since we no longer need to stop and restart Java daemons. By not switching back to Java mode, and being careful about what memory we touch between forks, we also keep the Zygote heaps much more stable, facilitating page sharing between the zygote and all its children. Under normal operation we should no longer allocate any memory in the zygote between forks. That applies to both the Java and C++ heap. This makes the zygote behave much more like the mental model many of us had assumed: It has nearly constant memory contents, which are copy-on-right cloned at each fork. This does not apply to the initial system server and webzygote forks, that are currently still handled differently. This includes 1. Add ZygoteCommandBuffer, and switch the argument parsing code to use it. This slightly reduces allocation and enables (3). 2. Support process specialization in the child, even when the arguments are already know, Leverages existing Usap code. 3. Add support for forking multiple child processes directly to the ZygoteCommandBuffer data structure. This directly uses the buffer internals, and avoids returning to Java so long as it can handle the zygote commands it sees. FUNCTIONALITY CHANGE: We now limit the total size of the zygote command, rather than the number of arguments. Initial performance observations: [ These are not perfect, since I'm comparing to numbers before I started. There may have been other moving parts, but they should be minor. ] System-server-observed launch latency: [Not the best metric, but easy to measure. In particular, this does not represent a significant reduction in application launch time.] Based on measuring the last 10 launches in a lightly used cf AOSP instance, the system server latency from requesting an app launch to response with the pid (which does not require the child to execute anything) went from an average of about 10.7(25) msecs to 6.8(9) and 7.9(16) in two tries with the CL. (The parenthetical numbers are maxima from among the 10; the variance appears to have decreased appreciably.) Dirty pages: The number of private dirty pages in the zygote itself appears to have decreased from about 4000 to about 2200. The number of dalvik-main private dirty pages went from about 1500 to nearly zero. Initially ART benchmarking service claim -1.88% in PSS. But this is not consistently repeatable. Drive-by fix: Call setAllowNetworkingForProcess on usap / native loop path. Bug: 159631815 Bug: 174211442 Test: Boots AOSP Change-Id: I90d2e381bada1b6c9857666d5e87372b6a4c1a70
2021-09-27[Havoc-11] SystemUI: add margin at the top of power menuAnushek Prasal
Change-Id: I2293b78722ba8fc2eb01d7831c1aab76b2d41310
2021-09-27[aospa][quartz] SystemUI: Fix pink progress animation color in expanded ↵Arne Coucheron
Wi-Fi tile For some reason, this is applying the completely wrong color. Following the code, it should show the default defined accent_device_default_dark color with an alpha of 0.6. Remove it to avoid the issue. Test: Turn off Wi-Fi. Expand Wi-Fi quick tile. Observe pink color on progress anim is gone. Change-Id: I2510a7cd90abb71e94b7698a7d28581517eb41b4
2021-09-22One more Build.FINGERPRINT to Build.DATE changemaxwen
Change-Id: I13dbf3d7f6587d3fcd6591cc0f861b34b6d5561c
2021-09-12SystemUI: Unblock gestural navigation on clearScreenshot()Timi Rautamäki
If user powers screen off when partial screenshot UI is visible, it will hide the UI but leave gestural navigation disabled. Also the partial screenshot UI would be visible when taking a new (normal) screenshot. Change-Id: Ie8138bb0b771ec0e21ceb48e52872c05c946d3b4
2021-09-12Merge tag 'android-11.0.0_r43' into ↵Kevin F. Haggerty
staging/lineage-18.1_merge-android-11.0.0_r43 Android 11.0.0 release 43 * tag 'android-11.0.0_r43': Use IntentFilter CREATOR directly for serializing ParsedIntentInfo Don't export HeapDumpProvider. Don't attach private Notification to A11yEvent when user locked Merge "BG-FGS-start while-in-use permission restriction improvement." into rvc-dev am: e51f884f6a Avoid locking profile task when it is already lock Improve ellipsize performance Fix side effects of trace-ipc and dumpheap commands DO NOT MERGE Add cross-user check for getDefaultSmsPackage(). Remove ParsedIntentInfo CREATOR Fix race condition between lockNow() and updateLockscreenTimeout Conflicts: services/core/java/com/android/server/policy/PhoneWindowManager.java Change-Id: I6a9a3e341be1fccaa568b875d4d9959d8f068794
2021-09-12Revert "Revert "Detects all activities for whether showing work challenge""Chirayu Desai
This reverts commit 2992a1dac025074726364d522cf9eac586f3d2c3. * This commit is present in android-11.0.0_r40, but not in android-11.0.0_r43 since that's on top of android-11.0.0_r39 * Just revert this to match what's shipping in r43 Change-Id: I8fe467db3bbf5aa40c10c4e1311829a733373279
2021-09-12Disallow click to partial screenshot right after screenshot is takenLuK1337
This change prevents triggering partial screenshot when trying to take full screenshot of protected window. Change-Id: I42a67ed1ee34edffccc6dacd17ba1dffc8d194b3
2021-09-07Fixed a crash in settings in tts engine selection screen.Deve
The crash happens when you try to uninstall current tts engine and install a different one and then engine name may be null. If engine name is null then you can't select new engine because of that crash, so that whole tts feature becomes unusable. Change-Id: Ie31a947eb6444242e8536db2dae1452ef74cc1ce
2021-09-07fixup! SystemUI: Show bluetooth battery level when availableLuca Stefani
The current update method is triggered in 2 cases: the SystemUI bluetooth controller changes device state the battery level changed intent is sent Sadly the current SysUI controller doesn't handle battery level changes, and is instead triggered in a few other cases ( ex audio state changed ). And since we use the data from SysUI controller for battery level from changes in the global bt controller we may have 2 different set of devices to work on. This could ( and does ) trigger race conditions. Instead add support battery level state changes within the SystemUI controller to work on the same data set. Change-Id: Ib57096fc2ebe151ac3860277a4db2319f230b514
2021-09-03fixup! AutoBrightness: Add support for one shot auto-brightness:Eamon Powell
* Remove unnecessary newlines * Fix incorrect indentation * Fix typo in comments Change-Id: I7d6071f4344605b3dcc3f1c33ac4390f1020e03f
2021-09-01Make battery clickable again in quick QS statusMarc K
Change-Id: I944b1122739754cc704f65039319fb260fcdbf1d
2021-08-23AutoBrightness: Add support for one shot auto-brightness:Cédric Bellegarde
- Only update auto brightness one time when screen is turned on. - Can be useful on devices where sensor is not accurate. Change-Id: I5187fdb7765550d4eaa999317ab95f487d44c8d8
2021-08-21Automatic translation importMichael Bestas
Change-Id: I32f53e719e6c274a78cfdafe71efb4a18d3f9f21
2021-08-14Merge cherrypicks of [15567411, 15567547, 15567447, 15567095, 15567367, ↵Android Build Coastguard Worker
15566999, 15567396, 15567397, 15567412, 15567413, 15567096, 15567448, 15567449, 15567097, 15567098, 15567450, 15567099, 15567398, 15567100, 15567621, 15567622, 15567468, 15567000] into rvc-qpr3-release Change-Id: I08d004280a03c61956f90404202a01d02ba05bc4
2021-08-14Make sure that only the owner can call stopVpnProfile()lucaslin
In stopVpnProfile(), it doesn't check if the caller's package name is the same as the given one, so any app has chance to stop the VPN profile of other apps. Bug: 191382886 Test: atest FrameworksNetTests CtsNetTestCases \ CtsHostsideNetworkTests:HostsideVpnTests Change-Id: Ib0a6e9ed191ff8c8bd55ce9902d894b6a339ace2 Merged-In: I254ffd1c08ec058d594b4ea55cbae5505f8497cc (cherry picked from commit f3072fcd46112bad7c5f6ddd4cc35d2c67f00d11)
generated by cgit v1.2.3 (git 2.25.1) at 2025-10-03 03:19:50 +0000