Stuck?

[francicco]

Hi,

It's days that Broccoli is running like this during step 3:

Why is that?
Thanks
F

[rderelle]

Hi,

the ML option is extremely slow.
NJ trees take about 0.2s while ML trees take about 10s to be computed, and Broccoli performs hundreds of thousands of phylogenetic analyses. Therefore, the computational time is multiplied by at least a factor 50 with the ML option (this factor increases with the number of species). This option shouldn't be used with more than 20 eukaryotic species or 60 prokaryotic species.

Also, I see that you are using 64 cpus. Broccoli analyses 1 proteome per cpu/thread -> there is no need to use more cpu than the number of species to be analysed. For 40 species, I would use max 20 threads (generally 8 cpus).

nb: I think both of these things are mentioned in the manual.

Hope that helps!
Romain

[francicco]

Hi Romain,

That's odd because I used ML option on previous runs, maybe I'm using too many cpus and it causing Broccoli to run ever slower.
Thanks, I'll adjust it
F

[francicco]

Hi Romain,

I confirm you, it was the number of CPUs that was slowing down the process.
Thank you very much!
Cheers
F

[drs]

@rderelle I had the same issue. Switching to using 8 threads fixed it, but made the step 1 and 2 slower. Is it possible to set a different (higher) number of threads for step 1, 2 than for step 3?