optimise Kv3_4.

[thorstenhater]

No description provided.

[Helveg]

• Benchmark: 12.6s/ms
• Changes: untested
  • Numerical
  • Functional

[thorstenhater]

Interesting, so it might not be the most impactful mechanism for your benchmark.

[noraabiakar]

This looks great! I think it can make quite a difference.

[Helveg]

Why did the function names etc change though, didn't they have biological meaning perchance?

[thorstenhater]

So, before we had a time constant $\tau = c/q$, but this was used only in $1/\tau$ position. So, I changed this
by defining $\rho = 1/\tau$ and folding some constants in. That saves two divisions (at least!) per update.

Can't the compiler do that for me? Yeah, probably, but why not give it a hand; this way we have the guarantee it's
done the way we want it.