Slow `co_forces`

[wcwitt]

Hi @cortner,

@casv2 and I have been making good use of the new committee stuff (thanks!), but we're having a problem with very slow co_forces. Maybe a type instability or something? Here are the basics, with files attached.

test.zip

julia> include("test.jl");
Time for forces:  0.007475964
Time for committee forces:  7.592369445

The committee forces are orders of magnitude slower, becoming prohibitively slow for cells of roughly 60 or larger. Any guesses?

Edit: I think my committee size is 50 here.

[casv2]

Thanks Chuck! Yeah this seems to really get worse for large cells, small cells up to 60 configs are OK.

[cortner]

Thank you - I'll look into it as soon as I can

[cortner]

partially completed by #53 - but I will leave it open since the performance is still poor and I don't understand the reason.