Scaling of timing on different processors and implementations

[ericagol]

• [x] starry version has quadratic scaling (with gradients) for large number of limb-darkening coefficients. Julia does not; is this an error in Julia or starry? profile Julia code to see where bottlenecks are.
• [x] julia v.0.7.0 is much slower than v.0.6.4
• [x] from the Julia command line, sometimes Julia code is much slower:

julia> @time timing=profile_transit_poly(trans,nb);
306.725521 seconds (460 allocations: 22.321 MiB, 0.00% gc time)

[ericagol]

Perhaps the problem with v0.7.0 is due to the factorial issue?

[rodluger]

@ericagol Did you fix this? The profile_transit_poly is running in the correct amount of time with 0.7.0 on the cleanup branch.

[ericagol]

@rodluger Yes, I replaced lfact with lfactorial, and fixed the other deprecations such as 1.+/-.

[ericagol]

@rodluger But, there are some other warnings that occurring that I need to fix before we close this issue.

[rodluger]

OK. I'm working on the cleanup branch as well, so there may be some merge conflicts we'll have to resolve when we try to push.

[ericagol]

@rodluger It also looks like there are some conflicts between the master and cleanup branches.

[ericagol]

@rodluger I've now updated the cleanup branch code and tests, and all of the tests seem to work now, both in Julia v0.7 and v1.0!

[rodluger]

Cool. Can you push? I've been editing those files as well.

[rodluger]

I'm going to set them up to run on Travis automatically.

[ericagol]

Yes, for some reason my first push didn't work; now I think it has (github asked me to log in again, for some reason).