Cut down on allocations

[1-Bart-1]

Currently:

examples/testing_stall_model.jl
julia> @time results_with_stall = solve(VSM_with_stall_correction, wing_aero_CAD_19ribs);
[ Info: Converged after 200 iterations
  0.008817 seconds (78.54 k allocations: 8.247 MiB)

Goal:
As close to zero as possible.

[ufechner7]

One option is to use MVectors and MArrays everywhere. This would reduce the allocations for Vectors and Arrays to zero. Limitation: Not good for vectors and arrays with more than 100 elements. In addition this requires to know the size of the vectors and arrays at compile time.

[1-Bart-1]

I already had to use StaticArrays to remove allocations somewhere. The problem is the mix of StaticArrays and Julia vectors we are getting now, it isn't very pretty. So yes, I would try using StaticArrays everywhere. Most arrays are under 100 elements and we know the size of the vectors and arrays on object creation.

[1-Bart-1]

Another option is to use AbstractArrays and AbstractVectors to let the user decide.

[ufechner7]

In the end there are two options:

• pre-allocated normal vectors and arrays
• StaticArrays

It would be good if the functions could support both, but that is more work to achieve then supporting only one of the two approaches.

According my experiance it is sometimes neccessary to use SVectors instead of MVectors to achieve zero allocations. Often it is quite clear which vectors might need to be modified later and which not. Finally, modifying SVectors in-place is also possible (but perhaps not very elegant).

[1-Bart-1]

I have preallocated the arrays in a lot of the code already, so I will continue with that method for now.