We’ve talked about ART before, and how it’s going to be some sort of replacement for Dalvik in the future, in order to make Android apps be much faster by getting translated to more native code that the processor can understand better and quicker. Right now, benchmarks show that apps compiled with ART (ahead of time, at installation), can be roughly twice as fast during operation, in relation to CPU usage, as you can see in the benchmark below (lower is better).
You’ll have to reboot your device afterwards, and it will take a while to reboot (10 minutes or so), because it needs to basically re-install every app, and compile it with ART at installation. The apps will also be a little bigger than before, because the size of the source code gets larger, which means when you install a new app using ART, it will take a while longer than it used to before. But I don’t think most people will even notice this, because apps vary from 1 MB to hundreds of MBs in size anyway, so when you’re installing some random app, you won’t feel that anything is different, since you might not know what was the original size anyway.
Although Google seems to have been working on this for the past 2 years, they have also recently acquired Flexycore, a company that was working on some similar technology, that promised to make Android apps even 3 times faster. Given that ART is still in experimental phase, and early days for its implementation, and that they now have the Flexycore engineers, too, I wouldn’t be surprised if Android apps compiled with ART would become even 3 times, or 4 times, faster than they are today.
I’m not expecting that to happen next year, though. If ART 1.0 arrives for the first time in a stable form in Android 5.0 (likely to happen), sometime next year, then it will probably have this 2x boost in performance. Further speed improvements will probably arrive in the next few years with ART 2.0, 3.0 and so on, until ART itself is mature enough, just like Dalvik is right now.
ART, together with the improvements we’ve already seen in KitKat for low-end devices, should help Android become just as fast, or even faster than any other mobile platform in the near future, and it should put Android even on $50 smartphones by early 2015 (that run well enough to be usable for most people), helping it reach that next billion users even faster.