This is not really an answer, but I feel so confined by the comment box ...
I do a fair amount of MPI programming, OpenMP too, but not CUDA and GPU stuff. I write mainly Fortran, some C++. I'm still using Emacs as my editor, and for the other things that Emacs does well. I use a separate parallel debugger (DDT, I've used TotalView in the past, more a question of which one is on the machine than which one I prefer) and a performance profiling tool called OPT (like DDT produced by Allinea Software).
I have looked, though not for a year or so, for plug-ins for NetBeans and Eclipse (former preferred, latter too Java-centric and too heavy these days) for parallel programming. What's out there is better for C++ than for Fortran. But I haven't yet come across any plug-in which has really made it far enough out of the research lab to be useful enough to make me change from the old ways.
I'll be as interested as you to see what other SOers recommend though right now it doesn't look very promising.