Proper way to access large matrix in a matrix list

General Usage Performance
11

It is, although that is not very flexible (they are constants afterall). Also, your code becomes hard to understand, as the result of a function depends on one or many things which are defined somewhere else.

Another alternative is to wrap all these parameters in a struct to pass only one parameter “down”. The Parameters package allows for a convenient constructor:

using Parameters
struct MyPars
   L::Int = 24
   mass::Float64 = 1.0
   basisN::Int = 4341
   kList::Vector{Int} = [0, 12]
end

newpars = MyPars() # defaults to those values
newpars = MyPars(L=48) # change something

function constructPsi_2!(coeffs_k, eigvecs_k, ψ; pars = MyPars()) # default 
    @unpack L, mass = pars  # "unpack" L and mass
    whatever = L * mass
    ...
    return ...
end

# call with:

constructPsi_2!(coeffs_k, eigvecs_k, ψ)

#or

constructPsi_2!(coeffs_k, eigvecs_k, ψ, pars = newpars)


#or


constructPsi_2!(coeffs_k, eigvecs_k, ψ, pars = MyPars(L=48))

Another thing: those type annotations in the function:

function constructPsi_2!(coeffs_k::OffsetArray{Vector{T},1, Vector{Vector{T}}}, eigvecs_k:: OffsetArray{Matrix{T},1, Vector{Matrix{T}}}, ψ::Vector{T})::Nothing where T

Don’t do anything for performance. They are useful for debugging and understanding the code better only. But specializing to an OffsetArray is certainly too much. You probably should at most use:

function constructPsi_2!(coeffs_k::AbstractVector{Vector{T}}, 
                         eigvecs_k:: AbstractVector{Matrix{T}}, 
                         ψ::AbstractVector{T}) where T

The Abstract there allows you to input a greater variety of vector types (such as offsets, slices, views).

2 Likes