You are welcome to see our results of using CUDA data compession algorythm (a noticeable 2x to 4x speedup is among them)!
We tackle not only complex problems but also complex software. We will assist you in speed, robustness and maintenance optimization. We will help you redesign bulky software systems with contorted and sloppy architecture to a punctilious and effective systems.
And we are having a good time implementing a.NET compression library for bzip2. No sci-fi, but the speed increased by a factor of 4. Not so bad yet.