Mateusz Baran

Yes, let's re-open it, maybe there is a better algorithm for this case.

Some people complain when StaticArrays sacrifices accuracy for speed, other people complain when it sacrifices speed for accuracy. StaticArrays can't meet everyone's needs. We are aware StaticArrays doesn't always use...

@thchr I've added some `@inbounds` to your code and it removed the allocation, roughly halving the runtime. Also note that for `\` we could have a more optimized Gaussian elimination...

We have the same issue with `getindex`: ```julia julia> (@SMatrix [1])[:,:,:] ERROR: BoundsError: attempt to access Tuple{} at index [1] ``` I guess it's just so rarely useful that no...

I think the plan is to transition to Manifolds.jl: https://github.com/JuliaManifolds/Manifolds.jl/issues/35 . There is more than these two functions that can be useful for manifold optimization.

The user-facing API for this is already implemented for `Simple` covariance. The only problem, I think, would be balancing efficiency and complexity of implementation. Both `cov(x::AbstractVector, c::CovarianceEstimator)` and `cov(x::AbstractVector, y::AbstractVector,...

Well, what I was saying is that existing methods already work when `p=1`, so you can get variance this way. I wasn't thinking about variance shrinkage methods that don't work...

Nice list, I would add one more: - [ ] [Nonlinear shrinkage estimation](https://arxiv.org/abs/1207.5322)

`repeat(Svals,1,50)` returning a dynamic array is actually quite reasonable because for arrays with more than roughly a few hundred elements static arrays are almost always slower than dynamic ones. `SMatrix{length(Svals),50}(repeat(Svals,1,50))`...

I wonder how much compilation time reduction is just due to not duplicating `U = A*U` in each `if` branch? Maybe we don't have to decrease performance for the 3x3...