Well today I worked a bit with OpenMP. It's not the first parallel stuff I've written, so the concepts are easy. What I like most about it is being able to incrementally parallelize your code. Its also nice that you don't have to deal with the lower level stuff of determining which threads do what. After working with MPI and UPC, I can say that OpenMP is much better. It'll probably end up being what is used for the client for multiprocessing. The good thing about it is the client can be developed sequentially and then we just go in and parallelize the bottlenecks or other code that just makes sense to do so. I don't know what direction the server team will take for parallel code. I suspect it will probably end up being MPI or a hybrid approach, combining OpenMP and MPI, which is a fairly popular approach. They have the additional complexity of not really knowing what server hardware to expect. Glad I don't have to worry about it.
Powered By Invision Power Board (http://www.invisionboard.com)
© Invision Power Services (http://www.invisionpower.com)
Blog system copyright © CJVJ