Re: OpenCL

[askok@...]

Yes, this and/or CUDA.

What would be required is a matrix handling system, where 
iteration code can be run on the GPGPU as Lua. In other 
words, just making a wrapper on top of OpenCL/CUDA API is 
clearly not enough.

This will be on my radar in 2009. But before doing any 
development on OpenCL they need to make it ready.

Hopefully, I will be able to integrate all this with 
Lanes, making multithreading with Lua even more powerful. 
:)

As to

Rob Kendrick wrote:
> I'm having trouble imagining a situation where there'd be a significant
> performance boost from running an interpreted general-purpose language
> on your GPU, though. They're really only any good at highly parallel
> workloads.

One scenario is meteorological data analysis, where 
non-linear operations are run for huge data matrices. In 
this case, we would be able to define the operations as 
Lua codelets, and run them parallel on the data. I guess 
only Lua of the current scripting languages (and Forth) 
would be small enough to effectively fit in on the 
CUDA/OpenCL. Even there, some limitations s.a. no 
recursive calls may be too much. But we'll see.

-asko



On Wed, 10 Dec 2008 10:47:10 +1300
 "Ian Bevan" <Ian.Bevan@Fronde.com> wrote:
> Hi
>
>
>
> I was wondering if anybody in the community has 
> considered porting LUA
> runtime to OpenCL? 
>
>
>
> http://en.wikipedia.org/wiki/OpenCL
>
>
>
> Thanks
>
>
>
> Ian