search

danielwilhelm / R-CS-ranks

R package implementing confidence sets for ranks
https://danielwilhelm.github.io/R-CS-ranks/
GNU General Public License v3.0
5 stars 1 forks source link

parallelise resampling #8

Open MrDomani opened 1 year ago

MrDomani commented 1 year ago

Currently whole code is sequential and spends most of time calculating the bootstrapped probability distribution, used later in hypothesis testing. We could quite easily speed things up by utilising more (all) CPU cores available. Implementing this change won't be difficult.

Using GPUs, on the other hand, would be difficult. I didn't see any user-friendly libraries on CRAN, that we could use. I guess we could use libraries directly supplied by CUDA, but I haven't done it before.

danielwilhelm commented 1 year ago

This is a good idea!

MrDomani commented 1 year ago

Note to yourself: avoid the overhead when num.cores == 1.