Skip to main content

Measuring Activity Startup Time

In recent talks I've given, as well as the Developing for Android series, I talk about the need to launch quickly, and about how to ensure that you're testing the right launch speed (cold start, like after reboot and (mostly) after killing a task, vs. warm start, which is way faster because the activity just has to be brought to the foreground).



Then someone asked me, quite reasonably, "So how do I get my app's launch time?"



Then I paused and wondered the same thing...



Whenever I've done this kind of benchmarking on framework code, I've had the ability to instrument exactly the bits I needed to. But how can non-framework developers get the information they need from just running the normal build?



Fortunately, this information exists, and has since API 19. So if you're running on any release later than 4.4 (Kitkat), you should be set.



All you have to do is launch your activity and look in logcat for something like this:






ActivityManager: Displayed com.android.myexample/.StartupTiming: +768ms







This information is output whenever an activity window is first drawn, after it goes through all of the startup stuff. This time includes the entire time that it took to launch the process, until the application ran layout and drew for the first time. This is basically the main information you need. It doesn't include the amount of time it took between the user clicking your app icon and the system getting ready to launch your activity, which is fine, because you cannot (as an app developer) affect that time, so there's no need to measure it. Instead, it includes all of the time it took to load your code, initialize your classes that are used at start time, run layout, and draw your app that first time. All of which is really what you want to measure, because that's what you can and should try to optimize.






There's an additional option to be aware of. The 'Displayed' time is automatically reported, to give you a quick measure of how long that initial launch took. But what if you are also loading in some other content asynchronously and want to know how long it took for everything to be loaded, drawn, and ready to go? In that case, you'll want to additionally call Activity.onReportFullyDrawn(), which will then report, in the log, the time between that initial apk start (the same time as that used for the Displayed time) and the time when you call the onReportFullyDrawn() method. This secondary time is a superset of the initial one (assuming you call it after the initial launch time, which is preferred), giving you the additional information about how long it took to do everything, including the follow-on work after the app was first displayed.






There is another way of measuring startup time which is worth mentioning for completeness, especially since it uses my favorite device tool, screenrecord. This technique involves recording the entire experience of tapping on your app's icon to launch it and waiting until your app window is up and ready to go.



First, start screenrecord with the --bugreport option (which adds timestamps to the frames - this was a feature added in L. I think):

$ adb shell screenrecord --bugreport /sdcard/launch.mp4



Then tap your app icon, wait until your app is displayed, ctrl-C screenrecord, and pull the file up onto your host system with adb pull:

$ adb pull /sdcard/launch.mp4



Now you can open the resulting video and see what's happening when. To do this effectively, you'll need to have a video player that allows you to step frame-by-frame (Quicktime does this, not sure what the best player with this feature is on other OSs). Now step through the frames, noticing that there's a frame timestamp at the top of the video window.



Step forward until you see the app icon highlighted - this happens after the system has processed the click event on the icon and has started to launch the app. Note the frame time when this happened. Now frame-step forward until you see the first frame that your application's full UI begins to be visible. Depending on your launch experience (whether you have a starting window, a splash screen, etc.), the exact sequence of events and windows may vary. For a simple application you'll see the starting window come up first, then a cross-fade with the real UI in your application when it's ready. You want to note the first frame where you see any of the real UI content of your app. This happens when your app has finished layout and drawn itself, and is now ready to be shown. Note the time at this frame as well.



Now subtract the two times ((UI displayed) - (icon tapped)); this is the full time that it took for your app to go all the way from the initial user tap to being drawn and ready. It is a superset of the "Displayed" log described above, since it includes time before the process launches and after that first rendering (when the system starts the cross-fade animation), but it is at least something that you can use for comparison purposes with other launches after you make things faster and want to see how much better it is.



As with any performance testing, it's good to try to run your tests in similar situations multiple times (including making sure you're testing 'cold start' as noted above), as various things can happen to vary the results on any one run.



Now that you know how to figure out your launch times, whichever approach you use, go make it faster.

Please.

Comments

Popular posts from this blog

Benefits Of Healthy eating Turmeric every day for the body

One teaspoon of turmeric a day to prevent inflammation, accumulation of toxins, pain, and the outbreak of cancer.  Yes, turmeric has been known since 2.5 centuries ago in India, as a plant anti-inflammatory / inflammatory, anti-bacterial, and also have a good detox properties, now proven to prevent Alzheimer's disease and cancer. Turmeric prevents inflammation:  For people who

Women and children overboard

It's the  Catch-22  of clinical trials: to protect pregnant women and children from the risks of untested drugs....we don't test drugs adequately for them. In the last few decades , we've been more concerned about the harms of research than of inadequately tested treatments for everyone, in fact. But for "vulnerable populations,"  like pregnant women and children, the default was to exclude them. And just in case any women might be, or might become, pregnant, it was often easier just to exclude us all from trials. It got so bad, that by the late 1990s, the FDA realized regulations and more for pregnant women - and women generally - had to change. The NIH (National Institutes of Health) took action too. And so few drugs had enough safety and efficacy information for children that, even in official circles, children were being called "therapeutic orphans."  Action began on that, too. There is still a long way to go. But this month there was a sign that

Not a word was spoken (but many were learned)

Video is often used in the EFL classroom for listening comprehension activities, facilitating discussions and, of course, language work. But how can you exploit silent films without any language in them? Since developing learners' linguistic resources should be our primary goal (well, at least the blogger behind the blog thinks so), here are four suggestions on how language (grammar and vocabulary) can be generated from silent clips. Split-viewing Split-viewing is an information gap activity where the class is split into groups with one group facing the screen and the other with their back to the screen. The ones facing the screen than report on what they have seen - this can be done WHILE as well as AFTER they watch. Alternatively, students who are not watching (the ones sitting with their backs to the screen) can be send out of the classroom and come up with a list of the questions to ask the 'watching group'. This works particularly well with action or crime scenes with