Lilith Orion Hafner

I went ahead and forced specialization as I suggested, resulting in this performance improvement and elimination of unexpected allocations caused by dynamic dispatch: ```julia julia> @b rand(100_000) sort!(_, alg=SortingAlgorithms.QuadSort) evals=1...

Would you share a specific benchmark that this outperforms the default algorithm on? I'm having trouble finding any on my hardware. CI on 1.0 is also failing.

I get similar results as you when running that function: ```julia julia> testsort(SortingAlgorithms.BlitSort) Success: 1; fraction: 0.9235749472202668 Success: 2; fraction: 0.8981071543150461 Success: 3; fraction: 1.0050755059840843 Success: 4; fraction: 1.0173432198064596 Success:...

> the function should be evaluated only once, because the second and all subsequent times, it is applied to the ordered collection, which gives a wrong expression `data` is never...

> This is a bit annoying, but AFAIK, there's just no good way to instruct Julia to create a stack-based array that elides the GC. Bumper.jl?

I don't think the solution in that issue fixes this because other pages which use the same `Last modified: {{ fill fd_mtime }}.` invocation are fine. I'm guessing it has...

Are you okay with removing it on all pages? Right now it is in footer, not the homepage specifically.

It's not standard but it is perfectly fine to make a PR from a master branch. See, for example, #38823, #39002, #39147, #39312, #39331, #39654, #39710, #39737, #39850, #39897, #40253,...

Thanks for working on this @cyanescent. We went ahead and merged #56423, incorporating the fixes that this PR brings.

It seems somewhat odd to require another package to pirate `LinearAlgebra.svd(::Base.AbstractMatrix)`.