[scala-language] The AMD JavaLabs team announced the open source release of Aparapi project for Java-OpenCL

[Qihui Sun]

"Aparapi is an API for expressing data parallel workloads in Java and a runtime component capable of converting the Java bytecode of compatible workloads into OpenCL™ so that it can be executed on a variety of GPU devices.

Originally an AMD product, Aparapi was released to open source on September 14, 2011. ReadGary Frost’s blog to learn more.

• To download Aparapi or contribute to the project, please visit the open source project site athttp://code.google.com/p/aparapi/.
• "

It seems similar with Stanford PPL's Delite,https://github.com/stanford-ppl/Delite(I heard EPFL joined this)

Might they collaborate.

--

Solomon

HUAWEI

Google+: Qihui Sun

[monster]

How does that compare to JavaCL and ScalaCL? Is it truly hardware-
independent, since it comes from one specific hardware maker?

[Gary Frost]

monster <skunkiferous <at> googlemail.com> writes:

>
> How does that compare to JavaCL and ScalaCL? Is it truly hardware-
> independent, since it comes from one specific hardware maker?
>
>

Aparapi supports any OpenCL 1.1 compatible runtime/device. The initial version
only supported AMD devices, the first patch added MacOS support and relaxed the
restriction. It will work with AMD, NVidia and Intel OpenCL runtimes.

Full disclosure, I am the Aparapi developer.

[Qihui Sun]

Well start,here is another excellent technology to utilize the GPGPU for Java:

http://www.ateji.com/px/java-on-gpu.html

2011/11/13 Gary Frost <frost...@gmail.com>

[Dave]

After installing the latest gpu driver with OpenCL 1.1 support
( http://developer.nvidia.com/opencl ）the aparapi examples detected my
gpu.

Would be nice if it somehow could work with Scala parallel
collections.
I have now 2 core cpu and 8 CUDA cores gpu.
Then I would have 10 logical processors :D