The multicore feature of `future::plan` has a delay in Radian

[crazyxiaoj]

Here's the R code snippet I was working with:

library(biomaRt)
library(Seurat)
library(ggplot2)
library(harmony)
library(magrittr)
library(dplyr)
library(future)

load("17_scRNA_base/scRNA_after_PCA.Rdata")
RhpcBLASctl::blas_set_num_threads(1)
options(future.globals.maxSize = 100 * 1024 ^ 3)
future::plan("multicore", workers = 16)

scRNA <- JackStraw(scRNA, num.replicate = 100, dims = 50)

When executing this code in base R, monitoring with htop -u shows that the multicore functionality from the future package engages smoothly right away.

However, in radian, it initially appears as though only a single core is active. After waiting approximately 3 minutes (tested with an Rdata file containing around 40,000 cells and about 500 MB in size), the multicore feature does start to work as expected.

Thank you for your time and for maintaining such a useful tool! I'm happy to provide more details or test any potential fixes.