Michael Helton
Michael Helton
Great thank you so much! Benchmarking always seems to be the issue. I will close this issue, but for some reason I am still seeing a discrepancy in my actual...
^^^^ For some reason I was thinking (hoping) I could get the constants to be evaluated at compile time if the sizes were known and fixed but I definitely misjudged...
Thanks for starting this discussion! I am very interested to see everyone's thoughts on this and open to whatever the consensus of the community is on this and would volunteer...
> We moved the log/exp functions from StatsFuns to a separate package LogExpFunctions a while ago and it seems development became more active This seems like a really great model...
Higher precision for noninteger orders is unlikely to be supported in the near future. You should check out an arbitrary precision library (e.g., `ArbNumerics.jl`). Also, I would be careful because...
I'd probably recommend keeping the `besselj0` and friends syntax. Scipy actually does [differentiate this ](https://docs.scipy.org/doc/scipy/reference/special.html#faster-versions-of-common-bessel-functions) and these functions are some of the most used. I'm not sure if a microbenchmark...
I was honestly a bit surprised too. I will probably open an issue over at [Bessels.jl](https://github.com/heltonmc/Bessels.jl) to continue this discussion there because this 3-4 ns difference holds across all the...
I went ahead and removed the instances of `@evalpoly` in the code base and changed to just `evalpoly`. Depending on how the discussion goes in #113 we can change back...
Sorry for continuing adding more commits (last one). I also added faster routines to compute the airy coefficients. These (I don't think) lend them self well to use `evalpoly` but...
I am unfortunately away from my desktop (Intel CPU) where I did this right this second but will run later. On my my laptop... ```julia julia> versioninfo() Julia Version 1.8.2...