GPU computing programming contests..

Well with all known good programming contests (Google Code Jam, ACM comp, Internation Informatic Olimpiad, Topcoder, University of Valladolid ..)
now you want to measure your proefficieny in not solving complex problems but also in mapping to extract all performance of GPUs.. well there are one going on.. and also some other coming?..

Let's see:

1.
There first I'm aware of is own compettions for course between peers:
ECE 498 AL : Programming Massively Parallel Processors
http://courses.ece.illinois.edu/ece498/al/
In that course a sort competition was started also showing the winner (kaatz) having a faster than CUDPP sort kernel.. I don't know how well it behaves compared to current fastest imp in CUDPP 1.1 or CUDA radixsort sample but I believe this new are now better..
http://courses.ece.illinois.edu/ece498/al/HallOfFame.html

Some year ago there was :
2.
1st Annual UMD GPGPU Programming Contest
supposedly coming every year and rewarding with Nvidia GPUs..
first was good trying to get fast sparse matrix vector and matrix matrix mul routines.. winner was among the fastest one implementation know at a time..
now there is CUSP..
note it has not worked this year..

3. Nvidia and TopCoder

Nvidia and Topcoder started a competition in September right before GTC and anounced the winers in GTC.. Right now the most impressive competition since is general (for everywhere) good prices (10000$) good learning pointers (CUDA webinars).. good hardware (gets tested on a GT200 Tesla with 4GB mem) and good problems..
First was about a labelling component problem..
I'm hosting PDF files for future reference.. and some example files (but not 400mb ones..)
Now is running 2 contest from 23 November to 7 December a graph problem..
Unique problem? is that no OpenCL right now.. but CUDA is more mature..

4. AMD
Also note that AMD made public in 58xx launch around 10 September it's intention of starting competitions in GPU programming.. Two kinds one like Nvidia one solving challenging problems and two is also interesting porting existing CUDA code to OpenCL and efficiently their hardware I supppose..
See slide in noticias3d.com review..

So to end my next post is talking about porting CUDA to OpenCL..