Peformance improvement : reduce engine init time 50%off at least #17192

lucky-chen · 2020-03-18T10:22:51Z

Hi, this pr try to reduce engine init time. In our data, init engine cost too much time: Android 300ms, iOS 380ms(most are mid-end and low-end machine)。

In our local test. Android reduce about 66.3% time，and iOS reduce about 46.6% time.

update:

Introduction change

Cause there are 15 files change, so Introduction to changes.

In fact , core modification all in shell.cc ,new method CreateAsync and CreateShellAsyncOnPlatformThread, other modification most are adapter code(android、iOS） and test。

engine (Core modification , also can work on other platform, just adapter Android/iOS now）
- runtime
  - shell.h
  - shell.cc
- test
  - shell_benchmarks.cc
  - shell_unittests.cc
adapter iOS
- runtime:
  - FlutterEngine.mm
  - FlutterEngine.h
- test:
  - FlutterEngineTest.mm
adapter android
- runtime:
  - FlutterEngine.java
  - FlutterJNI.java
  - platform_view_android_jni.h/.cc
  - android_shell_holder.h /.cc
  - FlutterNativeView.java
- test:
  - EngineLaunchE2ETest.java

ApiUsage

Android

final FlutterEngine engine = new FlutterEngine();
engine.initAsync( this,new EngineLifecycleListener() {
    public void onAsyncCreateEngineEnd(boolean success) {
        if (success) {
            engine.getDartExecutor().executeDartEntrypoint(DartEntrypoint.createDefault());
         } else {
            //balaba
        }
}});

iOS

 FlutterEngine* engine = [[FlutterEngine alloc] initWithName:@"my flutter engine"];
  [flutterEngine asyncRun:^(boolean success) {
    if (success) {
       [GeneratedPluginRegistrant registerWithRegistry:self];
    }else{
       //balaba
    }
  }];

googlebot · 2020-03-18T10:22:54Z

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here with @googlebot I signed it! and we'll verify it.

What to do if you already signed the CLA

Individual signers

It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
The email used to register you as an authorized contributor must also be attached to your GitHub account.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T10:22:54Z

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project (if not, look below for help). Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed (or fixed any issues), please reply here with @googlebot I signed it! and we'll verify it.

What to do if you already signed the CLA

Individual signers

It's possible we don't have your GitHub username or you're using a different email address on your commit. Check your existing CLA data and verify that your email is set on your git commits.

Corporate signers

Your company has a Point of Contact who decides which employees are authorized to participate. Ask your POC to be added to the group of authorized contributors. If you don't know who your Point of Contact is, direct the Google project maintainer to go/cla#troubleshoot (Public version).
The email used to register you as an authorized contributor must be the email used for the Git commit. Check your existing CLA data and verify that your email is set on your git commits.
The email used to register you as an authorized contributor must also be attached to your GitHub account.

ℹ️ Googlers: Go here for more info.

lucky-chen · 2020-03-18T11:07:31Z

@googlebot I signed it!

googlebot · 2020-03-18T11:07:34Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:07:34Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:10:52Z

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:10:52Z

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:54:58Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:54:58Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:59:38Z

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T11:59:38Z

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T12:12:27Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T12:12:27Z

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

googlebot · 2020-03-18T12:17:18Z

CLAs look good, thanks!

ℹ️ Googlers: Go here for more info.

liyuqian · 2020-03-26T17:52:32Z

@lucky-chen : flutter/tools/gn --runtime-mode=profile should generate an out/host_profile folder. Then ninja -C out/host_profile should build the shell_benchmarks binary. You can then run that binary to get benchmark numbers.

chinmaygarde

This is a good start and this model is definitely something we want to get to but there are some concerns we need to address first.

All Flutter engine APIs assume that the shell is ready to use after a valid reference to the same has been obtained via a call to Shell::Create. Now, depending on whether the Shell::Create call was made with the async_init_callback, all components that have access to the shell now have to wait for the callback to run. We want an API that doesn’t put the burden on the owner of a shell reference to know when said reference is safe to use. For example, if I were to vend a reference to the shell created using this callback specified to a component that needed its engine reference (by calling Shell::GetEngine), that component would have no idea if and how long to wait till it's safe to call GetEngine (or really, any other call on the shell). In fact, more worryingly, since the engine weak pointer has not been initialized yet, the call to GetEngine will now return a null weak pointer which will make the caller think that the shell has died already! As you can see, this API breaks some invariants in the Flutter Engine.

I think we can achieve the improvements without breaking the invariant described above though. Instead of adding just another argument to the Shell::Create call and vending a semi-instantiated shell, consider an API that just returns the shell itself in the callback. Now, shell initialization can begin but the reference will be vended to the caller only when it is safe to use. The caller can’t work on a semi-initialized engine because it won’t have a reference to it. The API would look something like. bool CreateShell(std::function<std::unique_ptr<Shell>()> shell_callback, … other arguments). When updating the iOS and Android embeddings, you can wait for the entire shell reference in the callback. This should not be an issue as it was not safe to use the shell reference before the callback anyway.

Also, please add a test for this in the shell_unittests target as that is the primary test target for this component.

chinmaygarde · 2020-03-27T07:50:38Z

@kangwang1988 I already left a review and see no updates or followup comments. Would you like me to clarify something?

kangwang1988 · 2020-03-27T08:33:05Z

@chinmaygarde
I mistakenly thought the refresh icon would trigger a double check over those resolved conversations. It's a misoperation over github. Please ignore it.

lucky-chen · 2020-04-03T10:00:13Z

hi,all, update finish .

benchmark

@liyuqian add benchmarks on shell_benchmark.cc. BM_ShellInitializationAsyncLockTime (LockTime) and BM_ShellInitializationAsyncTotalTime (LockTime + asyncTime)

Run on (8 X 2200 MHz CPU s)
2020-04-03 14:30:07
Benchmark                                     Time           CPU Iterations
---------------------------------------------------------------------------
BM_ShellInitialization                  4303584 ns    1033192 ns        689
BM_ShellShutdown                         900391 ns     377524 ns       1909
BM_ShellInitializationAndShutdown       5047434 ns    1225929 ns        574
BM_ShellInitializationAsyncLockTime     1186484 ns     996315 ns        648
BM_ShellInitializationAsyncTotalTime    4285084 ns    1014661 ns        694

ShellTestCase

@chinmaygarde add testcase on shell_unittests, adn all case run success

[ RUN      ] ShellTest.AsyncInitializeWithMultipleThreadButCallingThreadAsPlatformThread
[       OK ] ShellTest.AsyncInitializeWithMultipleThreadButCallingThreadAsPlatformThread (13 ms)

Introduction change

Cause there are 15 files change, so Introduction to changes.

In fact , core modification all in shell.cc ,new method CreateAsync and CreateShellAsyncOnPlatformThread, other modification most are adapter code(android、iOS） and test。

engine (Core modification , also can work on other platform, just adapter Android/iOS now）
- runtime
  - shell.h
  - shell.cc
- test
  - shell_benchmarks.cc
  - shell_unittests.cc
adapter iOS
- runtime:
  - FlutterEngine.mm
  - FlutterEngine.h
- test:
  - FlutterEngineTest.mm
adapter android
- runtime:
  - FlutterEngine.java
  - FlutterJNI.java
  - platform_view_android_jni.h/.cc
  - android_shell_holder.h /.cc
  - FlutterNativeView.java
- test:
  - EngineLaunchE2ETest.java

ApiUsage

Android

final FlutterEngine engine = new FlutterEngine();
engine.initAsync( this,new EngineLifecycleListener() {
    public void onAsyncCreateEngineEnd(boolean success) {
        if (success) {
            engine.getDartExecutor().executeDartEntrypoint(DartEntrypoint.createDefault());
         } else {
            //balaba
        }
}});

iOS

 FlutterEngine* engine = [[FlutterEngine alloc] initWithName:@"my flutter engine"];
  [flutterEngine asyncRun:^(boolean success) {
    if (success) {
       [GeneratedPluginRegistrant registerWithRegistry:self];
    }else{
       //balaba
    }
  }];

- add Shell::CreateAsync mode - supprot android && ios - add testcase for engine、Android、iOS - add benchmamrk on shell_benchmark.cc

liyuqian

Thank you for adding more benchmarks and tests!

A quick question: do you only expect the 50% init time reduction on add-to-app, or do you also expect such reduction on standalone Flutter Android/iOS apps?

Here's why I'm asking that:

While BM_ShellInitializationAsyncBlockTime and BM_ShellInitializationAsyncTotalTime are certainly nice to have, they don't tell a full story of your improvement. Specifically, those two benchmarks can't be run before your PR, so we couldn't use them to quantify your speedup.

Therefore, I tried to use our complex_layout__start_up test to compare the startup time before and after your PR, but I can't see a significant difference. You can run that test locally by running ../../bin/cache/dart-sdk/bin/dart bin/run.dart -t complex_layout__start_up --local-engine=android_profile inside flutter/flutter/dev/devicelab once you get your local engine compiled in profile mode.

I'd encourage you to write an independent benchmark like complex_layout__start_up to reveal your improvement. You can write a smaller PR to just include that benchmark. Smaller PRs also make it much easier for code reviews. As a result, your change will be approved and get merged faster 😄

liyuqian · 2020-04-06T23:04:46Z

shell/platform/android/io/flutter/embedding/engine/FlutterEngine.java

+    if (asyncInitCallback == null) {
+      Log.w(
+          TAG,
+          "initAsync: callback is null, you should care lifecycle, and called other api after initCallback called");


What is the Chinese version of this message? I'm having a little difficulty in understanding the English...

在异步模式中您需要关心初始化完成的时机，建议设置callback，并在callback触发后再调用其它api。😅

Maybe the following warning is a better translation: "initAsync: asyncInitCallback is null. Please be careful about the lifecycle and when the initialization is done. We recommend to set an asyncInitCallback, and call other APIs after such callback is executed."

lucky-chen · 2020-04-07T07:59:58Z

@liyuqian

add to app/Standalone

Just for add to app mode. At present, almost no one use standalone mode for existed apps.

benchmark:

Should compare BM_ShellInitializationAsyncLockTime with BM_ShellInitialization。
And must use new api to trigger the improvement , no effect with origin api.

engin.runAsync(...) //ios
engine.asyncInit(...) //android

BM_ShellInitialization ≈ BM_ShellInitializationAsyncTotalTime = BM_ShellInitializationAsyncLockTime + asyncTime

story

When using in an add2app scenario where flutter page will not be triggered until some user operations. Reduce time of app startUp, make homepage display as soon as possible

before
- click app icon -> mainthread locked by flutter entill init end (BM_ShellInitialization time )
- show homepage (show slowly, the display time is delayed )
- open second page(flutter or native or web)
after
- click app icon -> no lock wait,reduce much time when app startup (BM_ShellInitializationAsyncLockTime, other work with async mode asyncTime)
- show homepage (show quickly. The display time is unchange)
- open second page(flutter or native or web)

complex_layout__start_up PR

Will create another pr of asyncstartup benchmarks like complex_layout__start_up after the next few days。 But it dependens modify of this pr ...

liyuqian

Thank you @lucky-chen for the clarification!

Since this is specific to add-to-app, let's loop @gaaclarke into the code review as he used to work on add-to-app, and landed a similar PR that parallelize the engine initialization (#10182).

liyuqian · 2020-04-09T19:30:27Z

shell/platform/android/io/flutter/embedding/engine/FlutterEngine.java

+    if (asyncInitCallback == null) {
+      Log.w(
+          TAG,
+          "initAsync: callback is null, you should care lifecycle, and called other api after initCallback called");


Maybe the following warning is a better translation: "initAsync: asyncInitCallback is null. Please be careful about the lifecycle and when the initialization is done. We recommend to set an asyncInitCallback, and call other APIs after such callback is executed."

gaaclarke

I think this PR is perilous as-is because when creating the shell asynchronously there are no guarantees that the shell wont' be used anyway.

I think you what you've done here is good but could be better if we changed it a bit. Instead of creating 2 separate ways to make a shell, make creation always async and instead of relying on a callback to notify people when it's done, use a std::future. That way we can guarantee that the shell exists before it is used, no one can get it wrong.

gaaclarke · 2020-04-13T21:42:08Z

shell/common/shell.h

@@ -166,6 +168,13 @@ class Shell final : public PlatformView::Delegate,
      CreateCallback<PlatformView> on_create_platform_view,
      CreateCallback<Rasterizer> on_create_rasterizer);

+  static void CreateAsync(ShellCreateCallback callBack,


Needs a docstring. Please include the threads that will be invoking each callback.

gaaclarke · 2020-04-13T21:54:08Z

shell/platform/android/android_shell_holder.cc

+  init(empty);
+}
+
+void AndroidShellHolder::init(AsyncInitCallback holderCallback) {


Can we come up with a better name than holderCallback?

gaaclarke · 2020-04-13T21:56:11Z

shell/platform/android/android_shell_holder.cc

-      );
-
+  if (holderCallback) {
+    Shell::CreateAsync(


What ensures that shell_ is set before someone calls AndroidShellHolder::Launch?

gaaclarke · 2020-04-13T21:57:44Z

shell/platform/darwin/ios/framework/Headers/FlutterEngine.h

@@ -24,6 +24,11 @@ NS_ASSUME_NONNULL_BEGIN
 */
 extern NSString* const FlutterDefaultDartEntrypoint;

+/**
+ * the callback of engine async init mode


Full sentence documentation with proper punctuation please.

gaaclarke · 2020-04-13T21:58:05Z

shell/platform/darwin/ios/framework/Headers/FlutterEngine.h

+/**
+ * the callback of engine async init mode
+ */
+typedef void (^InitCallBackBlock)(bool);


/s/CallBack/Callback

gaaclarke · 2020-04-13T21:59:01Z

shell/platform/darwin/ios/framework/Headers/FlutterEngine.h

@@ -128,6 +133,10 @@ FLUTTER_EXPORT
 */
 - (BOOL)run;

+/** like run() method, but this method not block mainThread when init


Proper punctuation with complete English, please.

gaaclarke · 2020-04-13T21:59:52Z

shell/platform/darwin/ios/framework/Source/FlutterEngine.mm

- (BOOL)createShell:(NSString*)entrypoint libraryURI:(NSString*)libraryURI {
+- (BOOL)createShell:(NSString*)entrypoint
+           libraryURI:(NSString*)libraryURI
+    asyncInitCallback:(InitCallBackBlock)oc_async_init_block {


No abbreviations please.

gaaclarke · 2020-04-13T22:03:36Z

shell/platform/darwin/ios/framework/Source/FlutterEngine.mm

+              FML_LOG(ERROR) << "Could not start a shell FlutterEngine on asyncInitMode";
+              oc_async_init_block(false);
+            } else {
+              selfRef->_shell = std::move(shell);


What's to stop someone from invoking sendOnChannel:message: before _shell is set? How would they even know when it was safe to call it?

xster · 2020-04-15T01:33:42Z

shell/platform/android/io/flutter/embedding/engine/FlutterEngine.java

+        null);
+  }
+
+  public FlutterEngine() {}


This is dangerous since it bifurcates the meaning of having a constructed engine. Rather, create an asynchronous factory that calls back with a FlutterEngine instance when it's ready.

Generally, prefer composition over adding more functionalities and state to a single class.

xster · 2020-04-15T01:38:00Z

shell/platform/android/io/flutter/embedding/engine/FlutterEngine.java

    flutterLoader.startInitialization(context.getApplicationContext());
    flutterLoader.ensureInitializationComplete(context, dartVmArgs);

    flutterJNI.addEngineLifecycleListener(engineLifecycleListener);
-    attachToJni();
+    if (asyncInitListener != null) {


Since this is all about doing FlutterJNI.attachToNative() concurrently, you can just call that directly and not mix it into the state of the FlutterEngine instance. Do it outside this class.

xster · 2020-04-15T01:47:45Z

shell/platform/darwin/ios/framework/Headers/FlutterEngine.h

+/** like run() method, but this method not block mainThread when init
+ * @param block callbackBlock, called when async init end. must not be nil!
+ */
+- (void)asyncRun:(InitCallBackBlock)block;


Since you're adding an entirely new API, again for the sake of composable API and not mixing concepts together, you can actually just create a prepareVmAsync method or some such which gives you a callback when the Shell is created, at which point you can run run yourself.

Then we don't have to multiplex all the variants of run with an async variant.

lucky-chen · 2020-04-26T10:07:15Z

@liyuqian need help : build host_release/profile failed after update repo (Android/iOS build success)

ld: reference to bitcode symbol '___gxx_personality_v0' which LTO has not compiled in '___gxx_personality_v0' for architecture x86_64

Durning february, all of type build success. But seems to start from the march, after update repo, host always build failed

detail errormsg

$ ninja -C out/host_release -j 6
ninja: Entering directory `out/host_release'
[2262/3207] ACTION //third_party/dart/runtime:generate_version_cc_file(//build/toolchain/mac:clang_x64)
Warning: Could not get the most recent dev branch tag fatal: unknown refname: format lstrip=2

[2778/3207] LINK ./fml_benchmarks
FAILED: fml_benchmarks 
../../buildtools/mac-x64/clang/bin/clang++ -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.15.sdk -mmacosx-version-min=10.11.0 -flto -arch x86_64 -nostdlib++ -stdlib=libc++ -Wl,-dead_strip -Wl,-search_paths_first -L. -Wl,-rpath,@loader_path/. -Wl,-rpath,@loader_path/../../.. -Wl,-pie  -Xlinker -rpath -Xlinker @executable_path/Frameworks -o ./fml_benchmarks -Wl,-filelist,./fml_benchmarks.rsp  -framework Foundation -ldl -lpthread
ld: reference to bitcode symbol '___gxx_personality_v0' which LTO has not compiled in '___gxx_personality_v0' for architecture x86_64
clang-11: error: linker command failed with exit code 1 (use -v to see invocation)
[2780/3207] LINK ./fml_unittests
FAILED: fml_unittests 
../../buildtools/mac-x64/clang/bin/clang++ -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.15.sdk -mmacosx-version-min=10.11.0 -flto -arch x86_64 -nostdlib++ -stdlib=libc++ -Wl,-dead_strip -Wl,-search_paths_first -L. -Wl,-rpath,@loader_path/. -Wl,-rpath,@loader_path/../../.. -Wl,-pie  -Xlinker -rpath -Xlinker @executable_path/Frameworks -o ./fml_unittests -Wl,-filelist,./fml_unittests.rsp  -framework Foundation -ldl -lpthread
ld: reference to bitcode symbol '___gxx_personality_v0' which LTO has not compiled in '___gxx_personality_v0' for architecture x86_64
clang-11: error: linker command failed with exit code 1 (use -v to see invocation)
^Cninja: build stopped: interrupted by user.

Env

commit_id/engine_version: 4c8c31f
flutter doctor:

[✓] Flutter (Channel beta, v1.17.0-3.2.pre, on Mac OS X 10.15.3 19D76, locale zh-Hans-CN)
    • Flutter version 1.17.0-3.2.pre at /Users/chenpeihan/Library/flutter/flutter
    • Framework revision 2a7bc389f2 (4 days ago), 2020-04-21 20:34:20 -0700
    • Engine revision 4c8c31f591
    • Dart version 2.8.0 (build 2.8.0-dev.20.10)

[!] Xcode - develop for iOS and macOS (Xcode 11.3)
    • Xcode at /Applications/Xcode.app/Contents/Developer
    • Xcode 11.3, Build version 11C29
    ! CocoaPods 1.7.5.3 out of date (1.8.0 is recommended).
        CocoaPods is used to retrieve the iOS and macOS platform side's plugin code that responds to your plugin usage on the Dart side.
        Without CocoaPods, plugins will not work on iOS or macOS.
        For more info, see https://flutter.dev/platform-plugins
      To upgrade:
        sudo gem install cocoapods

Step

new dir
gclient sync
cd src/flutter -> checkout master
git reset --hard 4c8c31f
gclient sync
./flutter/tools/gn --runtime-mode release
ninja -C out/host_release -j 6

liyuqian · 2020-04-27T18:53:43Z

@lucky-chen : try ./flutter/tools/gn --runtime-mode release --no-lto. The --no-lto should only increase the binary size without impacting other performances. Meanwhile, --no-lto significantly reduces the compile time so I always used it in local development 😄

xster · 2020-04-27T23:41:52Z

You can also add your name to the AUTHORS file when you're ready for another look.

lucky-chen · 2020-04-30T05:11:43Z

hi, all. I split these changes into 3 PR. This way may be better to review and merge。

support engine create shell asyn future: support engine create shell with aysnc mode #18047
adapter to android (depends future: support engine create shell with aysnc mode #18047 )
adapter to iOS(depends future: support engine create shell with aysnc mode #18047)

after #18047 merged , i will submit the remaining 2 pr(Android/iOS)

@chinmaygarde shell.h/.cc modified according to your suggestion, look at #18047

bool CreateShell(std::function<std::unique_ptr()> shell_callback, … other arguments). When updating the iOS and Android embeddings, you can wait for the entire shell reference in the callback. This should not be an issue as it was not safe to use the shell reference before the callback anyway.

liyuqian · 2020-06-01T21:40:33Z

Shall we now close this in favor of #18047 ? (I understand that this PR includes more changes that will be submitted as follow-up PRs after the landing of #18047.)

auto-assign bot requested a review from stuartmorgan-g March 18, 2020 10:22

googlebot added cla: no labels Mar 18, 2020

lucky-chen force-pushed the master_async_init_pr branch from d1aabd7 to 9fd15dc Compare March 18, 2020 11:10

googlebot added cla: yes and removed cla: no labels Mar 18, 2020

googlebot added cla: no and removed cla: yes labels Mar 18, 2020

lucky-chen force-pushed the master_async_init_pr branch from 326a1c4 to 7c4d86b Compare March 18, 2020 11:59

googlebot added cla: yes and removed cla: no labels Mar 18, 2020

googlebot added cla: no and removed cla: yes labels Mar 18, 2020

lucky-chen force-pushed the master_async_init_pr branch 2 times, most recently from 44e8981 to 1ef74bf Compare March 18, 2020 12:17

chinmaygarde suggested changes Mar 26, 2020

View reviewed changes

liyuqian self-requested a review March 26, 2020 22:03

kangwang1988 requested review from kangwang1988 and chinmaygarde March 27, 2020 07:45

lucky-chen force-pushed the master_async_init_pr branch 2 times, most recently from 852fe7e to 4acbe5e Compare April 3, 2020 09:57

lucky-chen force-pushed the master_async_init_pr branch from 4acbe5e to 9e36331 Compare April 3, 2020 12:50

Reducing startup latency block time by 50% at least

3fc6fb4

- add Shell::CreateAsync mode - supprot android && ios - add testcase for engine、Android、iOS - add benchmamrk on shell_benchmark.cc

lucky-chen force-pushed the master_async_init_pr branch from 9e36331 to 3fc6fb4 Compare April 7, 2020 03:01

liyuqian reviewed Apr 7, 2020

View reviewed changes

liyuqian reviewed Apr 9, 2020

View reviewed changes

liyuqian requested a review from gaaclarke April 9, 2020 19:44

gaaclarke suggested changes Apr 13, 2020

View reviewed changes

xster reviewed Apr 15, 2020

View reviewed changes

kangwang1988 removed their request for review April 15, 2020 05:37

liyuqian mentioned this pull request Apr 16, 2020

Flutter cold start is slowly in some low-end Android devices flutter/flutter#35358

Closed

lucky-chen mentioned this pull request Apr 30, 2020

future: support engine create shell with aysnc mode #18047

Closed

liyuqian mentioned this pull request May 9, 2020

Setup default font manager after engine created, to improve startup performance #18225

Merged

lucky-chen closed this Jun 10, 2020

Peformance improvement : reduce engine init time 50%off at least #17192

Peformance improvement : reduce engine init time 50%off at least #17192

Uh oh!

Conversation

lucky-chen commented Mar 18, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Introduction change

ApiUsage

Uh oh!

googlebot commented Mar 18, 2020

What to do if you already signed the CLA

Individual signers

Corporate signers

Uh oh!

googlebot commented Mar 18, 2020

What to do if you already signed the CLA

Individual signers

Corporate signers

Uh oh!

lucky-chen commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

googlebot commented Mar 18, 2020

Uh oh!

liyuqian commented Mar 26, 2020

Uh oh!

chinmaygarde left a comment

Choose a reason for hiding this comment

Uh oh!

chinmaygarde commented Mar 27, 2020

Uh oh!

kangwang1988 commented Mar 27, 2020

Uh oh!

lucky-chen commented Apr 3, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

benchmark

ShellTestCase

Introduction change

ApiUsage

Uh oh!

liyuqian left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lucky-chen Apr 7, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lucky-chen commented Apr 7, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

add to app/Standalone

benchmark:

story

complex_layout__start_up PR

Uh oh!

liyuqian left a comment

Choose a reason for hiding this comment

Uh oh!

lucky-chen commented Mar 18, 2020 •

edited

Loading

lucky-chen commented Apr 3, 2020 •

edited

Loading

lucky-chen Apr 7, 2020 •

edited

Loading

lucky-chen commented Apr 7, 2020 •

edited

Loading

lucky-chen commented Apr 26, 2020 •

edited

Loading