Closed eix128 closed 6 years ago
AMD has released Aparapi for Java that makes so easy OpenCL and HSA development. http://developer.amd.com/tools-and-sdks/opencl-zone/opencl-libraries/aparapi/ Maybe it makes more easy to migrate OpenCL Here is some AMD APARAPI Examples: https://code.google.com/p/aparapi/wiki/UsersGuide
With aparapi , you write code once and runs on OpenCL or CPU Concurrent stack. That will be so good for Encog.
Thanks, I will look at aparapi. I've also considered adding a Linear algebra package, such as BLAS/CUBLAS that also contains a GPU version.
There is better library open sourced now. ArrayFire open sourced now :)
http://arrayfire.com https://github.com/arrayfire/arrayfire These , there is also java binding currently but i heard they make new java array fire binding. https://github.com/arrayfire/arrayfire_java
The advantage of array fire is , its also optimized for cuda and OpenCL
i don't know , it can be combination array fire + aparapi maybe. It depends on performance and scalability
At this point I really do not have the time to take the rather large undertaking of adding GPU support to Encog. More info here: http://www.heatonresearch.com/encog/encog_gpu.html
Hello , do you have any plan for OpenCL and HSA support ?