Android platform support

[moraaar]

What's new?

• Added Android platform support to PhysX 5.1.
• New build preset android-arm64-v8a added. Initial support focused on 64-bit arm8 architecture with NEON SIMD instructions.
• Support for generating PhysX as either static or dynamic libraries.
• New android platform readme.

Out of Scope

The following points are considered out of scope, they can be added in the future:

• Support for PhysX Snippets.
• Support for Omniverse Visual Debugger (OmniPVD) .
• Support for other architectures: arm7, x86, x86_64.
• Support for other SIMD instructions set: SSE.

Known Errors/Limitations

• The name of the bin folder for android will appear as bin\UNKNOWN. To fix this nVidia has to modify the CMakeModules package it's downloaded from packman when building PhysX. The file that needs to be modified is GetCompilerAndPlatform.cmake, adding the following lines to the if condition inside GetPlatformBinName function:

	ELSEIF(TARGET_BUILD_PLATFORM STREQUAL "android")
		SET(RETVAL "android.${ANDROID_ABI}")

• PhysX Systems that require a CUDA capable GPU are not supported in Android, for example particle system or cloth simulation.

Testing

• Built PhysX 5.1 successfully on android in debug, checked, profile and release using PX_GENERATE_STATIC_LIBRARIES set to both true and false. The right output binaries were generated in bin/android.arm64-v8a
• Built cmake install target successfully. It generated the correct output for Android under install/android-29 folder, only including necessary headers for Android.
• Runtime tested using Open 3D Engine (O3DE) with an early integration branch. The following video shows O3DE running PhysX 5.1 on Android. https://user-images.githubusercontent.com/27999040/203624867-6ed19905-d0dc-4dfe-943a-46e584fd9983.mp4

Signed-off-by: moraaar [email protected]

[CLAassistant]

All committers have signed the CLA.

[moraaar]

Android platform support has been submitted to the following PhysX fork where it will be maintained: https://github.com/o3de/PhysX

For more details see the following discussion: https://github.com/NVIDIA-Omniverse/PhysX/discussions/58