investigate selectobject julia method to see if there's any path to speedup there.

[ChrisBNEU]

Completing #1 would help this. We need to get a process in the EdgeAnalysis.selectobjects method in rms that can be parallelized using kernelabstractions.jl or some other julia gpu package.

[ChrisBNEU]

looking through the julia code, we have found 3 locations that would be good to try to parallelize with the gpu:

• calcbranchingnumbers
• calcradconsttermratios
• processfluxes
  these are taking a significant chunk of the overhead, and do have some fundamental array/matrix operations that we could put onto a gpu. To start, we will work with processfluxes, since this will run in a "vanilla" rmg execution (the other two are used only if you select certain options for your core enlargement).

Our concern is that there are a lot of memory write operations within these functions, so this might make any gpu speedup we attempt useless. we will try it first and post out results here.

[ChrisBNEU]

closing, we have a path for speedup in the three routines highlighted above. see issues #8 and #11.