A Faster Emulator with Better Hardware Support

[This post is by Xavier Ducrohet and Reto Meier of the Android engineering team. — Tim Bray.]

The Android emulator is a key tool for Android developers in building and testing their apps. As the power and diversity of Android devices has grown quickly, it’s been hard for the emulator keep pace.

Today we’re thrilled to announce several significant improvements to the emulator, including a dramatic performance upgrade and support for a broader range of hardware features, notably sensors and multi-finger input.

Added GPU Support

The system image we’re shipping today has built-in GPU support (Android 4.0.3 r2). With Android’s growing reliance on using the GPU to improve performance, the difference is significant. In the video below, the emulator is still interpreting ARM instructions; the performance boost is the effect of putting the GPU to work.

As a bonus, since we’re now supporting OpenGL ES 2.0, your OpenGL games can now run inside the emulator.

Please note that there are a lot of GPUs out there, and we haven’t tested all of them for this beta release, so let us know if you have feedback or encounter issues.

More Hardware Feature Emulation

The hardware features of mobile devices are a significant part of what makes them a unique platform for development, so we’re also pleased to announce that in addition to the camera support we added last year, it’s now possible to use a tethered Android device to supply inputs for sensors and multi-touch input.

We’re working on providing emulator support for more hardware features including Bluetooth and NFC.

Improved CPU Performance

We’ve also improved the CPU performance of the Android emulator. Hardware floating point operation has been available for system images since Ice Cream Sandwich (Android 4.0), allowing CPU operations to be emulated roughly twice as quickly.

Last week’s r17 developer tools release included x86 system images and host drivers (available through the SDK Manager), allowing the emulator to access the host CPU natively and offer significantly faster execution.

This video shows a CPU-bound application on two emulators running the same system image, one with virtualization, one without.

Building a modern emulator

Because the Android platform allows deep interaction between applications, and with system components, we need to provide an emulator with a complete system image. Our emulator virtualizes a complete device: hardware, kernel, low-level system libraries, and app framework.

Of course, the system being emulated typically has an ARM CPU; historically, we’d been emulating those instructions in software, and that worked OK until the advent of tablet support with additional animations and complexity in Android 3.0.

The missing pieces were the completion of Android x86 support, and the GPU support in last week’s release of SDK Tools r17. This works by funneling the OpenGL ES 2.0 instructions from the emulator to the host OS, converted to standard OpenGL 2.0, and running natively on the host GPU.

Conclusion

The Android ecosystem has a lot of devices in many different form factors. Developers need a good way of testing these apps without having to own everything out there and a fast, rich Android emulator is immensely helpful.

We hope that these new improvements will make the emulator a more useful tool in your development and testing, and look forward to improving it further for you.