Hello,
The open standard for parallel programming of heterogeneous systems is gaining popularity nowadays.
Nvidia (CUDA), ATI (Stream) and OpenCL (Open Computing Language) are supported by leading companies (see
http://www.khronos.org/opencl/)
There are many bindings to OpenCL (C#, java, fortran, python, delphi, etc.) but no Oberon ones.
So, I made a raw Oberon binding to OpenCL dynamic library and able to use GPU computing power for simple examples.
There are many kernels on the internet, (FFT, matrix, image,..) which needs to be adapted to Oberon enviroment.
I adapted and tested a few examples on WinXP32 with Nvidia 8500GT (2 GPU, 16 stream processsor, latest CUDA toolkit and driver )
graphics card using latest WinAos from SVN and on Ubuntu 10.04(lucid),x86_64 using LinuxAos rel-0.94. Component Pascal (Blackbox) version also tested.
There are many examples in OpenCL-SDK which can also be translated to Oberon.
There is a Matrix project in this OCP page also, so, which is better; make a new project for OpenCL or append it to Matrix project.
It is also possible to make a high level framework for OpenCL.
Are there any interest to extend and embed into Oberon, volunteer?
What is your opinion?
best regards,
- Necati.