CUDA is a framework for performing parallel computation on graphics processing units. My overall project is to write a software simulator of the a CUDA graphics card. The goal is to produce much more detailed profiling information, especially on how thread divergence and memory accesses create latencies in execution.
We met and decided that the best goal forward would be to write a preliminary CUDA simulator, and then use this simulator to optimize a piece of CUDA software. After that, a paper will be written, and submitted to a journal, probably iee or ieee letters. We originally thought a conference would be the way to go, but the conferences in the area all just had their submission deadlines pass.
Currently, I have the biggest parts of the simulator written, and I am now debugging and testing. After that is finished, we will integrate more realistic constants into our program. Then we move on to profiling, optimizing, and paper writing.