Hi, I noted that IBM Java SDK has been abadoned year ago for some reasons: https://gitweb.gentoo.org/repo/gentoo.git/log/dev-java/ibm-jdk-bin Well, the latest SDK provides quite interesting native GPU acceleration supporting the parallel stream API directly, so with lambda expressions you could utilize out-of-the box CUDA capable GPU, or you can use existing CUDA kernel written in C and bind it to your Java code. This is very interesting and game-changing feature, and I personally being investigating capabilities and performance of this. Could you guys consider pulling it back to tree? Thanks for info from Java team.
I'm not going to close this but we removed it partly because we're lacking manpower and that's only gotten worse since. Bumping these fetch-restricted packages is a pain. It's bad enough with just Oracle and right on cue, that needs bumping as of today.
I understand, I will try to prepare an ebuild on my own and will share it or put into my overlay.
Hey Ladislav Jech, did you succeed in making an ebuild? Regards
Hi Stig, actually not. I also thing about closing this ticket. IBM finally open sourced their JVM as OpenJ9, which consumes Eclipse OMR. On the other hand we have here Graal and different approach to generalization of VM to run any language. For my CUDA experiments I use https://github.com/bytedeco/javacpp-presets, where you can build your bridge to C++ libs and has CUDA binding available as well. On top of that there is open source project http://openjdk.java.net/projects/panama/ which integrates new way of native code integration. These and similar aspects are in my interests and I am not sure which way is better, still I think this ticket is obsolete at moment and won't be produced at all.