Posted by Mayank Jain – Product Supervisor, and Yasser Dbeis – Software program Engineer; Android Studio
Android builders have been telling us they’re searching for instruments to assist optimize energy consumption for various units on Android.
The brand new Energy Profiler in Android Studio helps Android builders by displaying energy consumption occurring on units because the app is getting used. Understanding energy consumption throughout Android units may also help Android builders determine and repair energy consumption points of their apps. They will run A/B assessments to match the ability consumption of various algorithms, options and even totally different variations of their app.
Apps that are optimized for decrease energy consumption result in an improved battery and thermal efficiency of the gadget, which suggests an improved person expertise on Android.
This energy consumption knowledge is made obtainable by way of the On System Energy Monitor (ODPM) on Pixel 6+ units, segmented by every sub-system referred to as “Energy Rails”. See Profileable energy rails for a listing of supported sub-systems.
The Energy Profiler may also help app builders detect issues in a number of areas:
- Detecting unoptimized code that’s utilizing extra energy than crucial.
- Discovering background duties which might be inflicting pointless CPU utilization.
- Figuring out wakelocks which might be retaining the gadget awake when they don’t seem to be wanted.
As soon as an influence consumption challenge has been recognized, the Energy Profiler can be utilized when testing totally different hypotheses to know why the app may very well be consuming extreme energy. For instance, if the difficulty is brought on by background duties, the developer can attempt to cease the duties from working unnecessarily or for longer intervals. And if the difficulty is brought on by wakelocks, the developer can attempt to launch the wakelocks when the useful resource shouldn’t be in use or use them extra judiciously. Then examine the ability consumption earlier than/after the change utilizing the Energy Profiler.
On this weblog put up, we showcase a way which makes use of A/B testing to know how your app’s energy consumption traits would possibly change with totally different variations of the identical function – and how one can successfully measure them.
An actual-life instance of how the Energy Profiler can be utilized to enhance the battery lifetime of an app.
Let’s assume you could have an app by way of which customers should buy their favourite motion pictures.
As your app turns into well-liked and is utilized by extra customers, you notice that a top quality 4K video takes very lengthy to load each time the app is began. Due to its giant dimension, you need to perceive its influence on energy consumption on the gadget.
Initially, this video was in 4K high quality in the very best of intentions, in order to showcase the very best film highlights to your clients.
This makes you suppose…
- Do you actually need a 4K video banner on the house display?
- Does it make sense to load a 4K video over the community each time your app is run?
- How will the ability consumption traits of your app change in the event you exchange the 4K video with one thing of decrease high quality (whereas nonetheless preserving the vivid look & really feel of the video)?
This can be a excellent situation to carry out an A/B take a look at for energy consumption
With an A/B take a look at, you may take a look at two barely totally different variations of the video banner function and select the one with the higher energy consumption traits.
State of affairs A : Run the app with 4K video banner on display & measure energy consumption
State of affairs B : Run the app with decrease decision video banner on display & measure energy consumption
A/B Take a look at setup
Let’s take a second and arrange our Android Studio profiler to run this A/B take a look at. We have to begin the app and fasten the CPU profiler to it and set off a system hint (the place the Energy Profiler will probably be proven).
Step 1
Create a customized “Run configuration” by clicking the three dot menu > Edit
Step 2
Then choose the “Profiling” tab and be sure that “Begin this recording on startup” and CPU Exercise > System Hint is chosen. Then click on “Apply”.
Now merely run the “Profile app startup profiling with low overhead” everytime you need to run this app from begin and fasten the CPU profiler to it.
Notice on precision
The next instance situations use all the app startup for estimating the ability consumption for this weblog’s function. Nevertheless you need to use extra superior methods to have even greater precision in getting energy readings. Some methods to attempt are:
- Isolate and measure energy consumption for video playback solely after a faucet occasion on the video participant
- Use the hint markers API to mark the beginning and cease time for energy measurement timeline – after which solely measure energy consumption inside that marked window
State of affairs A
On this situation, we run the app with 4K video taking part in and measure energy consumption for the primary 30 seconds. We are able to optionally additionally run the situation A a number of occasions and common out the readings. As soon as the System hint is proven in Android Studio, choose the 0-30 second time vary from the timeline choice panel and report as a screenshot for evaluating in opposition to situation B
As you may see, the common energy consumed by WLAN, CPU cores & Reminiscence mixed is about 1,352 mW (milliwatts)
Now let’s examine and distinction how this energy consumption modifications in State of affairs B
State of affairs B
On this situation, we run the app with low high quality video taking part in and measure energy consumption for the primary 30 seconds. As earlier than, we are able to additionally optionally run situation B a number of occasions and common out the ability consumption readings. Once more, as soon as the System hint is proven in Android Studio, choose the 0-30 second time vary from the timeline choice panel.
The full energy consumed by WLAN, CPU Little, CPU Large and CPU Mid & Reminiscence is about 741 mW (milliwatts)
Conclusion
All else being equal, State of affairs B (with decrease high quality video) consumed 741 mW energy as in comparison with State of affairs A (with 4K video) which required 1,352 mW energy.
State of affairs B (decrease high quality video) took 45% much less energy than State of affairs A (4K) – whereas the decrease high quality video gives little to no visible distinction in perceived high quality of the app’s display.
On account of this A/B take a look at for energy consumption, you conclude that changing the 4K video with a decrease high quality video on our app’s dwelling display not solely reduces energy consumption by 45%, additionally reduces the required community bandwidth and may doubtlessly additionally enhance the thermal efficiency of the units.
In case your app’s enterprise logic nonetheless requires the 4K video to be proven on the app’s display, you may discover methods like:
- Caching the 4K video throughout subsequent runs of the app.
- Loading video on a person faucet.
- Loading a picture initially and solely load the video after the display has totally rendered (delayed loading).
The general energy consumption numbers offered within the above A/B take a look at situation might sound small, nevertheless it reveals the methods that app builders can use to successfully A/B take a look at energy consumption for his or her app’s options utilizing the Energy Profiler in Android Studio.
Subsequent Steps
The brand new Energy Profiler is accessible in Android Studio Hedgehog onwards. To know extra, please head over to the official documentation.