Performance of a key step in Strang Splitting method

Man, I had no idea Julia was capable of this kind of magic. I’d only heard of reinterpreting in that fast inverse square root from Quake III. Just to confirm, are you getting the exact same results for ϕ in all of these cases? I ask because I, having run only my reconstructed f1 and f2 as I don’t have 1.6 installed, don’t get the same ϕ in both cases. Do you mind sharing your f1 and f3 for completeness sake*?

Oh and as far as discourse goes I’m not sure which post to mark as the “answer” as I’d like to thank everyone for helping out.

*and also because I am not so experienced and haven’t seen the @threads macro used before so might mess it up…