Web Seminar: Programming GPUs Beyond CUDA

[xman]

GPU/CUDA programming is easy if we ignore the performance, or even the correctness of the program. It becomes tough when the performance is critical, one has to optimize very hard on the specific hardware. Fortunately, GPU hardware performance improves drastically every 2 years. Unfortunately, the performance is not portable across different generations of GPUs.

Prof Chen from Tshing Hua University is proposing MapCG, a MapReduce framework as a resolution to the portability problem.

Check out the details of the seminar in the following link:

http://www.gpucomputing.net/?q=node/5277

[Chad Brewbaker]

A paper on MapCG:

http://www.cse.ohio-state.edu/~agrawal/788-au10/Papers/Nov4/mapcg-hetero.pdf

I don't think Map/Reduce is a panacea. Sort, scan, reduce, and MatMul are still king. They are right that a clean high level API for these... say in Ruby ;)  would be nice. 

[xman]

Algorithms that map well into map reduce framework would shine. Otherwise, we are really trading the performance for portability. Whether this makes sense, I think the maintenance cost, development cost really come into play. Just as we use Ruby for higher productivity, trading off some performance.