If you call lu on an SMatrix, you get a static LU factorization—cool: using StaticArrays using LinearAlgebra A = SMatrix{3, 3}(rand(3, 3)) lu(A) # StaticArrays.LU # L factor: # 3×3 LowerTriangular{Float64, SMatrix{3, 3, Float64, 9}}: # 1.0 ⋅ ⋅ # 0.706739 1.0 ⋅ # 0.64896…

[image] maxkapur: But sometimes, you want to preallocate space for the L and U factors. Preallocation doesn’t make sense for SMatrix — there is no “allocation” cost in the first place. So what is your motivation?

For small matrices, won’t my_LU \ b benefit from having my_LU as an SArray for the same reason that A \ b does? (Or is the latter assumption also incorrect?)

lu(smatrix) already returns L and U factors that are backed by SMatrix. What more do you want? (Note that StaticArrays already has optimized methods for triangular solves with SMatrix storage.)

How to preallocate a static array for use with `lu!`

General Usage

stevengj Split this topic November 3, 2022, 2:28pm 5

2 posts were split to a new topic: How to preallocate LU factorization

Topic		Replies	Views
How to preallocate LU factorization Numerics linearalgebra	3	352	November 3, 2022
How to reduce large memory allocation by lu! with sparse matrices Performance question , memory-allocation , ode , sparse	8	1546	July 17, 2021
In place lufact for sparse matrices and prellocation for lufact output Numerics performance , linearalgebra	1	635	September 26, 2017
Replacing normal arrays with static arrays New to Julia linearalgebra	6	1433	August 31, 2019
Solving linear system without allocating Numerics	19	643	August 9, 2025

How to preallocate a static array for use with `lu!`

Related topics