RenderScript is a component of the Android operating system for mobile devices. It is an API for intensive computation using heterogeneous computing. It allows developers to increase the performance of their applications at the cost of writing a greater amount of more complex code. It provides the developer three primary tools: A general purpose compute API across different system computing hardware, a compute API similar to CUDA, OpenCL or GLSL, and a familiar C99-derived language. It can also be used for 3D graphics. Android 4.2 added new capabilities to script intrinsics, such as Blend and Blur; as well as ScriptGroups which allow you to chain together related RenderScript scripts and execute them with one call.
Most recently, Google added FilterScript, which is a subset of RenderScript that allows developers to write their image processing operations in FilterScript using the standard RenderScript runtime API, but within stricter constraints that ensure wider compatibility and improved optimization across multi-core CPUs, GPUs, and DSPs. FilterScript is a less precise in terms of numeric datatype precision, and more cross device compatible subset of RenderScript – and should not be mistaken for a RenderScript replacement technology.
- The API was first published with the Android Honeycomb API release
- It consists of a "computing" API
- Accessed using a C-based scripting language
RenderScript is designed to always run on the various Android platforms regardless of hardware type. Its goal is code portability for various architectures and not performance.
RenderScript portability depends upon device-specific drivers: a basic CPU-only driver is provided for every device, while there exist some specific chipset-provided RenderScript drivers that enable GPU usage (e.g. Qualcomm specific drivers, which are provided in the
libRSDriver_adreno.so Android library).
The API is designed to accommodate tasks that can be efficiently split and run concurrently on the underlying hardware which may be only known at the runtime of the application.
As of Android 4.2, RenderScript has been expanded to run on the GPU in addition to the CPU on supported systems.
Simplify development as much as possible, where possible steps are automated to avoid glue code.
It's these trade-offs that separate RenderScript from the existing approaches on the device, such as Dalvik or the Android Native Development Kit. They should be thought of as different tools intended to solve different problems.
- RenderScript cannot yet express on-chip inter-thread communication (known as local memory in OpenCL, and shared memory in CUDA).
- RenderScript cannot yet express hardware-implemented 2D and 3D lookups with bilinear interpolation (known as texture in CUDA, and image read in OpenCL).
- "Android 4.2 APIs". Retrieved 2013-03-20.
- Marchetti, Alberto (2016). RenderScript: parallel computing on Android, the easy way (1st ed.).
- "Jelly Bean - Renderscipt Performance". Retrieved 2012-11-27.