Just download the project, test the different algorithm. Also, read about the theory because it can be complex.
And then, start to develop.
I add you as a contributor ;-)
Welcome
Krys
-----Original Message-----
From: "Sarma Tangirala" <tvssarm...@gmail.com>
Sent: Friday, June 10, 2011 2:05pm
To: cl...@googlegroups.com
Subject: [Intro] Hello
A very interesting project will be to port the Duane Merril scan and radix-sort from CUDA to OpenCL.
The challenge is that it should work too on AIT GPU and any CPU. You can even optimize it for ATI !
http://www.cs.virginia.edu/~dgm4d/
http://code.google.com/p/back40computing/
Just to let you know I think that the next step is to produce a version 1.0 of CLPP.
So, what is missing :
1 - There is a bug with the sorting algorithm, only for the ATI GPU. The graphic cards is locked due to synchronization problem. I haven't got any time to fix it !
It is the issue here : http://code.google.com/p/clpp/issues/detail?id=4
2 - Implementing Bitonic sort for the GPU. In fact, at least for the first step I thinkg that we can use the code from http://www.bealto.com/gpu-sorting_intro.html
I have been in touch with Eric Bainville and he allow us to integrate it in CLPP. I think that we only have to do some adaptation to match the CLPP library structure, interfaces etc...
So, if you are interested to help us, I think it is a good way to start.
But now it is up to you.
Thanks