nf-core
Number of cores calculation for Trimgalore
Description of the bug
As discussed in this issue, the calculation for actual number of cores used by Trimgalore is convoluted. The current calculation is present in this code. But, the actual core usage seems to be more than the number provided by task.cpus when the parameter --gzip is active with python3 and pigz available.
Breaking it down and drawing parallel to the comment by @FelixKrueger in the same thread, in the trimming section,
- cutadapt uses cores provided by the
-jargument - trim_galore: 1 (high)
- write to pigz stream: 1 (low)
- however, pigz would also use
-jcores
This would mean, at least 2 * j + 1 cores for the trimming process.
The same would extend to the validation stage where due to pigz, the calculation would include additional j cores than the prior calculation.
Command used and terminal output
Relevant files
No response
System information
No response
This is how the nf-core module (here) is implemented please open an issue there with your comments to update the module. Thanks!
