5-arg `mul!` for continuously coupled operators?

[jlchan]

Is there a mul! operation specialized for SummationByPartsOperators.UniformNonperiodicCoupledOperator? If not, I'd be happy to help add it if that's useful.

[ranocha]

I think I should have implemented everything in terms of mul!, e.g.,

SummationByPartsOperators.jl/src/coupling.jl

Lines 424 to 439 in 4854524

	function mul!(dest::AbstractVector, cD::UniformCoupledOperator, u::AbstractVector, α=true)
	N, _ = size(cD)
	@boundscheck begin
	@argcheck N == length(u)
	@argcheck N == length(dest)
	end
	@unpack D, meshgrid, coupling, der_order = cD
	if coupling === Val(:continuous)
	mul!(dest, D, meshgrid, coupling, der_order, u, α)
	scale_by_inverse_mass_matrix!(dest, cD)
	else
	mul!(dest, D, meshgrid, coupling, der_order, u, α)
	end
	dest
	end

If that doesn't work, please let me know! PRs are always welcome, of course.

[jlchan]

Sorry, I didn't realize that was already there. I was looking for methods for UniformNonperiodicCoupledOperator and saw

julia> methods(mul!, (SummationByPartsOperators.UniformNonperiodicCoupledOperator,), SummationByPartsOperators)
# 0 methods for generic function "mul!":

Looking more closely, I think the issue is that I need a 5-argument mul!, which I hope should be easy to add. Would you like me to go ahead and submit a PR for that?

[ranocha]

Ah, I see. Yeah, the case of 5-arg mul! is more complicated for this case. The CGSEM operators are easier to write down at the level of M * D, so I extracted the multiplication by the inverse mass matrix into another function. To get full 5-arg mul!, this needs to be re-written. If you have time to spare, I would love to review a PR from you for this 👍

[jlchan]

Sounds good. I'll try implementing it later this week.