Multithreaded variance computation

[ummel]

I imagine @SebKrantz has considered this, so this question is mainly out of curiosity:

Is there a reason that multithreading is not possible for fvar() and fsd()?

For example, using Chan's parallel algorithm to combine single pass results computed on separate parts of x?

[SebKrantz]

Hi, yes, it is possible with the unstable algorithm available when stable.algo = FALSE. I just haven't done it yet. This is also not the default so it was not urgent. You can already get up to 4x speedup using the unstable algorithm.

[ummel]

Thanks, @SebKrantz. I agree it is not a priority. And I assume that multithreading across columns (as opposed to within) would be an easier parallelization strategy for fvar(), anyway.