John F. Gibson

My feeling is that MileCranmer is right, that having up-to-date benchmarks against just open languages is better than having them held up, going back multiple Julia versions, in order to...

The `rangemax` illustrates the point well but for the benchmarks I feel like a clearer connection to a widely-familiar application would be better (IMHO). I've tried to illustrate this point...

Yes... ```julia function rungekutta4(f, x, N, Δt) Δt2 = Δt/2 Δt6 = Δt/6 t = 0.0 for n = 1:N s1 = f(t, x) s2 = f(t + Δt2, x...

Timings on that Julia code and two C implementations: one that passes f to rk4 by a function pointer, another that inlines f by defining it inline prior to the...

I'm coming around to `argmax`. Maybe it's worth having both `userfunc_argmax` and `rungekutta4_lorenz`, the latter using libraries for rk2 and user funcs for lorenz.

I think there are some issues with these codes | heaps(n), n=1:4 | code | |-------------|--------| | 1, 3, 12, 60 | kalmarek Julia | | 0, 3, 12, 60...

@kalmarek aha! thank you. (regarding returning the sum of the first entry). I ran what's currently posted above and got heaps(1:5) == 1, 3, 12, 60, 360. There were some...

If you look towards the bottom of the notebook there's an "implement Numba and Dedalus codes" in the Room for Improvement to-do list. I'd be grateful for a pull request...