Daniel Karrasch

Not to insist, but to learn something: I read https://docs.julialang.org/en/v1/manual/performance-tips/#Be-aware-of-when-Julia-avoids-specializing as it will specialize since in our case, the `Function` arguments are called within `_hermkron!`, not just passed as an...

Could well be me (who suggested it) at the time. I remember looking it up not so long ago.

How about this now? The signature could be relaxed again once somebody writes sparse kron-kernels, but for now I think this is not too bad.

This looks good to me. I'm curious about a couple of things? 1. What's the runtime impact on, say, addition of `::Diagonal + ::Tridiagonal`? 2. Does this have any latency/TTFX...

Regarding removing methods: I checked and found, e.g., this method: ```julia @commutative function (+)(A::Bidiagonal, B::Diagonal) ``` What would happen, if we removed this and the similar below this? This is...

I guess this is expected behavior, since the outer `pmap` already employs (by default) all workers, so there are no ones left to distribute to. Check the `?pmap` docstring. However,...

Looks like this is ready to go then? Perhaps cudablas (and maybe others?) could use the same trick, if wanted? @maleadt

Sounds more like we need ```julia _realtype(::typeof(abs2), T) = _realtype(T) _realtype(::typeof(abs), T) = float(_realtype(T)) ``` ?

Indeed, I was confusing `abs` and `norm`. `norm(1)` returns `1.0`.

Which Julia version is this? The currently released version of LinearMaps.jl is 3.11.3, not 3.11.2. With v3.11.3, I can't reproduce on v1.6, v1.10 and v1.11-rc. Could you please show the...