I am just starting into the world of Cuda. I am going to be converting a financial time series modelling application from .Net to run most of its calculations on the GPU.
The intended machines it would run on are the Amazon EC2 Cluster GPU instances.
After reading Cuda by Example and working my way through a number of the available webinars, I think I have figured out a high level outline of how I can architect things.
That said, experience has taught me that before going off and starting to write code, it is always wise to try to ensure that what I am planning on doing is, in fact, a very good way of doing it. Along these lines, I figured I would post a thread to seek some perspective.
Being new to the Cuda world and to this forum, I am uncertain how to best seek this sort of high level design feedback. Is that sort of thing done on this forum or is it, instead, the type of input that is best sought by finding a consultant? How have other members who have traveled this path before me calibrated their design ideas to be reasonably optimal?
Any perspective is appreciated.