Java的当前OpenCL实施是否提供跨平台支持以及要在项目中使用的稳定性？

Question

Background: Our school entered in a competition where my team (based on 3 people including me) is assigned the task to do navigation and image processing. It's an year long project but I want to get started as soon as possible and see what are the options I have. University and other firms are funding in this research as well. I am no computer science major, we are building an engineering application that highly depends on navigation and image processing for basis of intense calculation purposes on top.

Question: I understand we will be needing algorithms and smart data structure choices to cut down the processes and split the work in threads to harness the power of the device. Anyhow, I have been reading a lot regarding OpenCL implementations for Java that I found namely, LWJGL, JOCL, JavaCL. The problem is, they all do the same thing but some does it better in their own unique ways. I am looking for something simple with fair amount of community and cross-platform support. Because, the microcontrollers we will buy most probably run Linux (I am thinking Arduino and SMARTGPU).

Anyways, this is very early so I need some pointers as to where to go from here. I will probably need to look up some guides, tutorials, and manuals as well to understand more on OpenCL.

Answer 1

I have been using JavaCL for nearly a year now and have found it to be easy to use and reliable. The one bug that I did find was fixed within 24 hours (admittedly a very simple bug). I did try using JOCL but found their API to be a little difficult to use compared to JavaCL. I have no experience with LWJGL, but would point out that LWJGL is more general as it is intended for games.

If you decide to use Java, then I would suggest starting with JavaCL.

However, I would suggest that you consider OpenCV as it provides a lot of image processing functions that have already been heavily optimised. It may not be as fast as using OpenCL and a fancy embedded GPU but it could cut your development time.