Re: Octave and CUDA

[mpender]

I haven't heard of Clover or AMP yet; I'll have to research them.  I did look
into Matlab and the parallel processing library, but the parfor
implementation did not understand how to handle distributing my code across
multiple processors. 

I also looked into OpenCL previously and concluded that NVIDIA is putting a
lot of time and resources into developing CUDA as a computing platform but I
haven't seen the same level of progress with OpenCL development.  When it
does reach maturity then it will work on the NVIDIA products too, so there
isn't any down side for me to become familiar with the NVIDIA hardware
strengths and weaknesses.  

The NVIDIA compiler is 'free' in that it does not cost anything, but not
free in the "open source software" sense.  Although this is not really
surprising in view of the level of implementation-specific details regarding
NVIDIA's products required to make the NVCC compiler generate fast code. 
I'm using Linux since I prefer it over Windows as a development environment. 
I don't have to spend a small fortune continuously updating my development
tools for what is essentially a personal interest project.

I understand if the NVCC compiler license means that it cannot be
distributed with Octave, but IMHO that shouldn't prevent development of
tools to enhance Octave and tap the awesome number crunching power available
for people willing to accept the license terms.  Teraflops of computing
power on the desktop should not be dismissed lightly.

- Mike



--
View this message in context: 
http://octave.1599824.n4.nabble.com/Octave-and-CUDA-tp4660998p4661100.html
Sent from the Octave - General mailing list archive at Nabble.com.