Closed DanielVandH closed 1 week ago
I think the idea is that layout
represents the desired memory layout that has already been computed? Does this lead to issues somewhere?
It doesn't lead to any issues that I've seen, I only noticed it while implementing the two methods elsewhere and saw the difference, so I wasn't sure if it was intentional or not.
https://github.com/JuliaLinearAlgebra/ArrayLayouts.jl/blob/a77697d30fb4c954f05a4bb101e6c2e2f8e13a39/src/factorizations.jl#L382-L386
Should the call to
_chol!
be passingMemoryLayout(A.data)
instead of simplylayout
? Something similar is done forreversecholesky
herehttps://github.com/JuliaLinearAlgebra/MatrixFactorizations.jl/blob/d3f30bb62fc282e9a3873a858f2823339888e143/src/reversecholesky.jl#L113-L117
which I think is more correct?