rgt-hint footprinting threads using?

[Rqafu]

Hi HINT teams

When I'm processing my ATAC-dataset with the step footprinting, my collegues complain that I took all the threads from server for quite some time (more than 30 min per sample) and it seems there's no parameter in this command to limit the cpu usage...? It would be super nice to have a extra paramter for this command to limit cpu usage :)

Best, Run

[lzj1769]

Hi,

Thanks for your feedback. Currently, in the footprinting step, we didn't limit the number of cores. I will try to add this in our next release. :)

For differential footprinting analysis, you can set the max. number of cpus by using --nc

best, Li

[AlexBlais74]

Hello I was searching the issues regarding this --nc parameter and I arrived here. The OP and Li's reply are a bit confusing and imply a number of things. I should mention that I run RGT on a large cluster with jobs managed by SLURM. Also, I am running version v1.0.1 of rgt-hint

Here are my questions: -is the rgt-hint footprinting command multithreaded? Will it use all computer cores we give it access to?

-how about the rgt-motifanalysis matching command? Is it multithreaded? And if so, how is this controlled?

-what is the exact meaning of the --nc X parameter, and what is the default value if the parameter is skipped? I have been assuming it means "distribute the job to X processes" and the fact of not specifying this option would make it run on a single CPU core, but the discussion here seems to imply more like "distribute the job to NO MORE THAN X processes"

-how exactly is the number of cores detected? I am asking because on our SLURM-managed cluster, different nodes have cpus with varying numbers of cores. When I ran rgt-hint footprinting --nc 2 with --nodes 1 and --n-tasks 2 in the SLURM options, the program reported:

32 cpus are detected and 2 of them will be used...

I find this intriguing since SLURM will in fact give me access to no more than 2 cores. My guess is that rgt-hint detects the chip architecture (and reports it somewhat inaccurately as 'cpus' instead of 'cores') instead of the exact number of truly available cores. The majority of nodes on our cluster indeed are 2 x Intel E5-2683 v4 Broadwell chips with 32 cores.

Thanks in advance for the clarifications. Alex