Android GPGPU Testing Prototypes
Update January 2015: Please note, that I have released a library for OpenGL ES 2.0 based GPGPU for Android and iOS systems -- ogles_gpgpu.
This repository contains some of the prototypes that I developed during my master's thesis -- "Parallel Computing for Digital Signal Processing on Mobile Device GPUs." More information about my thesis and the PDF is available on my website.
These prototypes test different algorithms (e.g. image convolution, Hough transform) with different GPGPU technologies and give report about the execution time of individual processing steps. The following technologies are assessed:
- OpenCL (projects are prefixed with "Cl")
- OpenGL ES 2.0 shaders (projects are prefixed with "Gl")
- Android RenderScript (projects are prefixed with "Rs")
For details and results, please refer to the project page on my website.
Unfortunately, the code is not very well documented because of the general time pressure during writing a master's thesis. Nevertheless, I hope that this helps people who want to get started with GPGPU on mobile devices.
The code has been tested with a Nexus 10 running Android 4.2.1. Please note that OpenCL support is not available on most devices (see p. 20 of my master's thesis) and requires manual setup (see this post and that post).
- ClAudio -- OpenCL-driven simple real-time sound synthesis
- ClImageProc -- OpenCL-driven image convolution and Hough transform with kernel generator
- CvMarkerDetect -- Marker based augmented reality prototype based on OpenCV and partly accelerated on the GPU via OpenGL ES 2.0 shaders
- GlImageProc -- OpenGL ES 2.0 shader driven image convolution with kernel generator (uses Java API)
- GlImageProcNative -- OpenGL ES 2.0 shader driven image convolution, thresholding and Parallel Coordinate space based Hough transform (uses NDK API)
- RsImageProc -- Android RenderScript driven image convolution (limited functionality, crashes)
Licensed under BSD 2-Clause license. See the file