Wanting to optimize this function and I know I have some un-needed allocations. Just wondering what the best choice is to fix a few lines: function singleAtomDistances!(distMat::Array{Float64,3},crystal::Crystal,LJ::LJ,centerAtom::Vector{Float64}, centerType:: Integer, loopBounds::Vector{Int64}) # …

[image] lancejnelson: newAtom = neighboratom + [i, j, k] # I know this is a problem This is a poster child for using StaticArrays.jl to represent your Cartesian coordinate vectors. Also, your distMat indices are in the wrong order for locality. indices = sort([iNeighbor,centerType]) Th…

[image] stevengj: This is a poster child for using StaticArrays.jl to represent your Cartesian coordinate vectors. But the values of the vector change each iteration.

[image] stevengj: Beware the Matlab/Python habit of using arrays for every tiny calculation. I know I’m still very Pythonic in my coding and trying to break that habit. But I also want elegant code so … trying to find the balance.

[image] lancejnelson: But the values of the vector change each iteration. If you have a loop with an integer, i and do i += 1 on every iteration, do you worry about the value of i changing on every iteration? The situation with StaticArrays is the same. If you do for neighboratom in aType…

Thank you… and I’m with you… I just don’t understand how the compiler knows the length of @SVector[i,j,k], but doesn’t know the length of [i,j,k]

PS. In general, it will be a lot easier to help if you if you provide minimal runnable examples. That means providing all necessary functions (or stripping/stubbing them out), providing sample data, etcetera. Please read: make it easier to help you

[image] lancejnelson: I just don’t understand how the compiler knows the length of @SVector[i,j,k], but doesn’t know the length of [i,j,k] The basic thing to keep in mind is that Julia’s compiler (like many compilers) specializes code mainly based on the types of objects, not based on their v…

Right… The type of the data is so key in Julia. I’ve got to remember that. Thank you. Sometimes I wish I hadn’t learned Python first because it trains you to not think about types at all. My knee jerk reaction when I think I have too many allocations is to pre-allocate the array and then modify th…

See also: What's the deal with StaticArrays? - #2 by stevengj Whats the difference between a regular array and an array from StaticArrays

[image] lancejnelson: distMat[indices...,1] += 4. * 1.0/2.0 * 1.0/r^6 What about the splatting. I think this is allocating too. I mean indices isn’t very long so I suppose I can just type the indices out distMat[indices[1],indices[2],1] += 4. * 1.0/2.0 * 1.0/r^6 but is splatting a bad ide…

Optimizing function with vector reassignments

New to Julia

stevengj October 4, 2024, 9:42pm 10

Topic		Replies	Views
Assigning in place to avoid allocations New to Julia	5	647	June 16, 2021
Remove allocations from splatting? General Usage question , memory-allocation	18	695	July 19, 2023
Vector notation, intermediate value and allocations New to Julia	8	528	June 16, 2022
Allocation with StaticArrays New to Julia memory-allocation , staticarrays	19	791	September 15, 2021
Performance improvements for simple molecular surface caclulation Performance	15	748	September 14, 2020

Optimizing function with vector reassignments

Related topics