Andreas Noack
Andreas Noack
I looked at the main repo and it seems that they have changed that struct quite a few times so if we really want to match it in Julia, we'd...
This has been reported a couple of times. It only happens when Julia has been built from source but we are we are trying to figure out how to fix...
Not yet, unfortunately. I suspect we'd need to do separate binaries for MKL. cc @staticfloat
Would probably be best to add tests as well here. Is the PR otherwise ready?
I'm not sure if this is an issue. I'd think that the reproducibility is only for a fixed element type.
Do you really want to differentiate `rand`? I guess you are aware but generally I don't think sampling algorithms give any guarantees about differentiability unless you explicitly sample via the...
> my `test(x)` function above is a deterministic smooth function of `x` It happens to be the case and it would be a bit odd if this wasn't the case...
A pathological example for the fun of it ```julia myrand(rng::Xoshiro, x::Normal) = isodd(reinterpret(UInt64, x.σ)) ? rand(rng, x) : (rand(rng); rand(rng, x)) ``` would still satisfy all our current requirements for...
Yeah. I can see that so I'd be fine with adding such a method as long as it has a comment with a reference to the method in Random necessitates...
I can confirm this. I also get ``` julia julia> using PyPlot julia> figure() Figure(PyObject ) julia> 1.0 ERROR: syntax: invalid numeric constant "1.0" ``` I have ``` julia julia>...