Java Sumatra project for OpenCL

I found an interesting article for Java's Smatra project.
http://pc.watch.impress.co.jp/docs/column/kaigai/20130826_612591.html

I was interested in using OpenCL in Java 3 years ago, at that time there was no such projects, but  2 years ago, a solid project, Sumatra project to support OpenCL in Java seems started by Oracle/AMD.
OpenCL is very important standard to support heterogeneous hardware environment including GPU, CUP, etc. OSX of Apple is now based on LLVM (Since Snow Leopard) and supporting OpenCL. So This move of Java is actually following Apple.

Project Sumatra

This primary goal of this project is to enable Java applications to take advantage of graphics processing units (GPUs) and accelerated processing units (APUs)--whether they are discrete devices or integrated with a CPU--to improve performance.
We expect the initial focus to be on the Hotspot JVM, enabling code generation, runtime support and garbage collection on GPUs. Once basic capabilities are in place in the JVM, we will also examine how to best expose GPU support to application and/or library developers, initially leveraging the new Java 8 Lambda language and library features. As the project progress, we may identify challenges with the Java API and constructs which may lead to new language, JVM or library extensions that will need standardization under the JCP process. While the Java language is an obvious focus, we also anticipate that this project will provide guidance for other JVM-hosted languages such as JavaScript/Nashorn, Scala and JRuby.
You can follow the project and get involved on the sumatra-dev mailing list.

Related projects and material

• Aparapi - API for data parallel Java
• Rootbeer - a compiler that makes it easier to program GPUs with Java
• Arrays 2.0 - a JVM Language Summit talk by John Rose

This Project is sponsored by the HotSpot Group.

This is quite interesting direction, also Java 8's support of Lambda(closure) seems associated with this parallel computing based on GPGPU or OpenCL approach.

I installed two projects in Windows. but it looks like Intel only supports Windows OS(except xeon). In order to use this on Linux, we may need to install AMD's SDK. It seems usable for Intel's graphics core in the CPU(like i5 etc).

1) https://code.google.com/p/aparapi/
2a) http://developer.amd.com/tools-and-sdks/heterogeneous-computing/amd-accelerated-parallel-processing-app-sdk/
2b) http://software.intel.com/en-us/vcsource/tools/opencl

These are other related projects.

http://openjdk.java.net/groups/hotspot/

http://openjdk.java.net/projects/mlvm/index.html

http://download.java.net/openjdk/jdk7/

Here is a qupte from jdk8 project. It actually seems related to Sumatra project..
 http://mreinhold.org/blog/jdk8-preview

JDK 8 Developer Preview

2013/09/09 10:39:26 -07:00 

The JDK 8 De­vel­oper Pre­view (a.k.a. Mile­stone 8) builds are now avail­able!
This mile­stone is in­tended for broad test­ing by de­vel­op­ers. We’ve run all tests on all Or­a­cle-sup­ported plat­forms and haven’t found any glar­ing is­sues. We’ve also fixed many of the bugs dis­cov­ered since we reached the Fea­ture Com­plete mile­stone back in June.
The prin­ci­pal fea­ture of this re­lease is Pro­ject Lambda (JSR 335), which aims to make it eas­ier to write code for mul­ti­core proces­sors. It adds lambda ex­pres­sions, de­fault meth­ods, and method ref­er­ences to the Java pro­gram­ming lan­guage, and ex­tends the li­braries to sup­port par­al­leliz­able op­er­a­tions upon streamed data.
There are, of course, many other new fea­tures, in­clud­ing a new Date and Time API (JSR 310), Com­pact Pro­files, the Nashorn JavaScript En­gine, and even some anti-fea­tures such as the re­moval of the Per­ma­nent Gen­er­a­tion from the HotSpot vir­tual ma­chine. A com­plete list of new fea­tures is avail­able on the JDK 8 Fea­tures page.
If you’ve been watch­ing JDK 8 evolve from afar then now is an ex­cel­lent time to down­load a build and try it out—the sooner the bet­ter! Let us know if your ex­ist­ing code doesn’t com­pile and run cor­rectly on JDK 8, if it runs slower than be­fore, if it crashes the JVM, or if there are any re­main­ing de­sign is­sues in the new lan­guage and API fea­tures.
We’ll do our best to read, eval­u­ate, and act on all feed­back re­ceived via the usual bug-re­port­ing chan­nel be­tween now and the end of Oc­to­ber. After that we’ll grad­u­ally ramp down the rate of change in order to sta­bi­lize the code, so bugs re­ported later on might not get fixed in time for the GA re­lease.

Other Java 8 related info:

http://www.infoq.com/news/2013/06/jdk8-almost-feature-complete

https://blogs.oracle.com/thejavatutorials/entry/jdk_8_documentation_developer_preview

This kind of direction should be considered in Dart as well.
This will affect both of client and server. Since JavaScript seems going to support OpenGL like architecture, Dart need to handle it also. But server side support would be more natural fit for high performance computation using OpenGL.

In fact, I was surprised to know the direction of Java 8 was shifted to this direction. Previously it was aiming to support of modules etc, but the focus seems changed quite  a lot.