Jutho

I will keep it open and merge it at a suitable time; currently my git history is somewhat messed up due to other changes locally.

Thanks for opening this issue. Part of my motivation for `KrylovKit` was to go beyond the hassle of having to define a linear map first, but more so because I...

The `@time` macro in Julia Base should give you a hint as to total time spent in garbage collection. There will certainly be other factors contributing to this time difference...

Another suggestion might be to enable multithreading. Due to the way KrylovKit works, it won't collect the different vectors that span the Krylov space into a matrix (because they might...

Yes, KrylovKit currently uses the Schur decomposition algorithm from BLAS, and thus only works with BLAS numbers. With GenericSchur.jl that could be circumvented, and I welcome PR requests to do...

`Q` comes directly out of a Schur factorization from Lapack (`hseqr!`) so that is indeed very strange. Do you have a minimal working example?

There shouldn't be. Is your matrix an actual matrix (i.e. `AbstractMatrix') or is it a function. Does the function have any side effects? Does it not mutate the incoming vector?

That's strange: ```julia julia> λ, V, info = eigsolve(A, x0, 2, :LR) (Complex{Float64}[0.153975+0.0im, 3.20255e-8+0.0im, 1.86884e-8+0.0im], Array{Complex{Float64},1}[[9.53135e-16+0.0im, -1.30201e-15+0.0im, 1.27389e-16+0.0im, -1.45716e-16+0.0im, -2.09105e-16+0.0im, -4.76706e-16+0.0im, -6.8875e-16+0.0im, -4.49207e-16+0.0im, 6.70572e-18+0.0im, -2.50869e-16+0.0im … -0.0327677+0.0im, -0.00120309+0.0im, -0.00388521+0.0im, 0.0133307+0.0im,...

Ok, the random initial vector triggers it indeed. The error is not surprising: ``` * = 1: two adjacent blocks were too close to swap (the problem * is very...

@nrontsis , yes of course I should do that at the very end. Thanks for the pointers, @haampie How urgent is this prolbem @nrontsis ; I will definitely fix it...