GPU acceleration

So I just found this post http://blogs.nvidia.com/blog/2013/09/22/gpu-coming-to-java/
Im very excited about this as it removes the requirement for java to use external librarys and that we will be able to make games without having to use LWJGL OPENGL JGL etc

I don’t get it. How is this different from what our bindings to OpenCL/OpenAL/OpenGL offers?

You will still have to use all these libraries in the future for stuff which wish to explicitly use the GPU (3D graphics and so on).
The big BUT is, that with this technologie all the other Java apps will profit from the GPU without the developer even knowing.

I think you’ve got a bit confused about what that article is about.

You’ll still need all the binding libraries to make games.

Cas :slight_smile:

What I mean is not for games , but when designing code you could specifically assign tasks to the GPU without having to use any external librarys.

It’s possible but doubtful that the general purpose Oracle JVM will receive this sort of hardware acceleration for certain algorithms under the hood. Don’t hold your breath though.

Cas :slight_smile:

Normal game/app code would gain nothing from gpu acceleration.

Actually… there are a few things that would benefit from it in my game - for example the painter’s algorithm sort in the sprite engine has to sort 20-30k sprites per frame and a 2x speed increase would be welcome. There are possibly other places which could benefit from a little bit of super-parallelism like the collision manager but that won’t “just work” from JDK code.

Cas :slight_smile:

Lot of game code could benefit from gpu acceleration but not automatically. It’s just like automatic vectorization, sounds good in paper but does not benefit usually at all.

More the to point. Why would you want to? I want to manage what both the CPU & GPU is processing.

(EDIT: For clarity…I’m only talking about game runtimes here.)