Daniel Karrasch comments

Results 291 comments of


                                            Daniel Karrasch

ugly special-casing of `AbstractQ` for concatenation (`cat`)

I'm not sure we can get any further than #1500. In concatenation, anything that is not an `AbstractArray` is treated as a scalar of size `(1,)`. These methods were needed...

when should inv(factorization) be another factorization?

I agree. In JuliaLang/LinearAlgebra.jl#633 I think I was mislead by the specifics of the SVD.

Symmetric eigen memory usage for eigenvectors.

IIRC then that commit doesn't change the behavior, it just avoids creating copies of arrays. Do you observe a difference in array size compared to, for instance, v1.11?

Symmetric eigen memory usage for eigenvectors.

I'm not sure I'm following. This is a random matrix with a random spectrum, so it could be that there is one less eigenvalue in the unit disk. Or, maybe...

Symmetric eigen memory usage for eigenvectors.

Ah, now I spotted the difference. Thanks!

Symmetric eigen memory usage for eigenvectors.

How does that express itself in terms of performance? I guess the `summarysize` reflects the fact that the array was larger initially, but it was allocated in either case. I...

`diag` with a `Val` band index

This may increase latency a bit, but should decrease allocations and potentially speed-up reading from the bands?

Optimize `copytrito!(B, A, 'L')`

What are the improved runtimes?

Optimize `copytrito!(B, A, 'L')`

> The third comment of [Reference-LAPACK/lapack#1183 (comment)](https://github.com/Reference-LAPACK/lapack/issues/1183#issuecomment-3748656316) gives a workaround. Shall we use it before LAPACK is updated? Yes.

Optimize `copytrito!(B, A, 'L')`

Seems ready, doesn't it? I checked that there is no overhead in the wide rectangular case for the LAPACK case.