Re: Project proposal for GSoC.

[Daniel J Sebald]

On 02/26/2012 09:05 PM, Jordi Gutiérrez Hermoso wrote:
> 2012/2/26 Daniel J Sebald<address@hidden>:
> > It would be interesting to see how GPUs could be included in Octave.
>
> And I think that before we get razzle-dazzled by all the shiny things
> that GPUs can do, we should tackle the more fundamental problem that
> they can't do these things freely.
>
> At least there are free compilers. What good is a free numerical
> environment largely compatible with Matlab if it can't compute freely
> when using GPUs?

I'm not sure what you mean.  The "free" part of Octave doesn't pertain 
to what the software is run on.  I had to buy my computer.  If I wanted 
to buy a compiler instead of gcc and compile Octave, I could, right? 
There's nothing in the license I know of that restricts use of Octave in 
that way, and there shouldn't be.

You ask "what good is...?" which I interpret to mean "Who would use 
this?"  In other words, is there a big enough user base to make the 
effort worthwhile?  My point is that because OpenCL seems to be an 
emerging standard that many are adopting (and in some cases dropping 
their own approach)

http://en.wikipedia.org/wiki/OpenCL
http://en.wikipedia.org/wiki/OpenGL
http://www.khronos.org/opencl/

that yes there could be a big user base.

We aren't talking about including libraries or hunks of code that are 
proprietary in Octave.  I don't completely understand OpenCL's standard, 
but it's a case of writing code with extensions to the C standard and 
that code would remain open as part of Octave.  The standardization 
makes it an interesting twist on what Ben was getting at.

The bigger issue is how to incorporate something like that.  Not easy. 
Surely integrating it into Octave core is not a task for a mere mortal. 
 And compiling it as a plug-in of sorts might be just as daunting.

Dan