Parallelization of short alignments

119 views
Skip to first unread message

Kaján Győző

unread,
Nov 29, 2018, 7:55:03 AM11/29/18
to ra...@googlegroups.com
Dear Alexis,

Following your advice I moved (back) to RAxML from Exabayes (thank you for the detailed answer!) for the analysis of short alignments (few hundred residues long). The alignments might be short, but contain often numerous (thousands of) taxa. Thus I would like to use some parallelization. I did love that RAxML 8.X could infer the requested number of trees parallel on different CPU cores. You also suggest such parallelization in the manual, and I did try to achieve such an analysis setup. But I am not sure I succeeded, it seems, that RAxML-NG did use fine-grained parallelization and split my short alignment.

So my aim is to set up RAxML-NG for parallel tree searching: if I need e.g. 100 trees, 1 core infers 50 trees and the other core the other 50 trees parallel. At the end RAxML-NG should pick the best-scoring topology from the 100 as the final ML tree.
How should I set up RAxML-NG? In most cases I would like to use a single node but multiple cores. (Should MPI be used for this purpose?)

And a theoretical question:
How many replicates one should infer?

Attached are the files resulting from two analyses: a small one and a big one. For the big one I could attach a job statistics figure too, which shows that all 5 cores are working 100% all the time.

Best regards and thank you in advance,
Gyozo

Ps: Just read today's digest about the "number of free parameters" error message. Had the same error with the big one, that is why the --force mode is ON.
bash_small_RAxML_NG_test.txt
bash_big_RAxML_NG_test.txt
big.raxml.log
small.raxml.log
big_core_usage.jpg

Alexandros Stamatakis

unread,
Nov 30, 2018, 2:29:34 AM11/30/18
to ra...@googlegroups.com
Dear Kajan,

> Following your advice I moved (back) to RAxML from Exabayes (thank you
> for the detailed answer!) for the analysis of short alignments (few
> hundred residues long). The alignments might be short, but contain often
> numerous (thousands of) taxa. Thus I would like to use some
> parallelization. I did love that RAxML 8.X could infer the requested
> number of trees parallel on different CPU cores. You also suggest such
> parallelization in the manual, and I did try to achieve such an analysis
> setup. But I am not sure I succeeded, it seems, that RAxML-NG did use
> fine-grained parallelization and split my short alignment.
>
> So my aim is to set up RAxML-NG for parallel tree searching: if I need
> e.g. 100 trees, 1 core infers 50 trees and the other core the other 50
> trees parallel. At the end RAxML-NG should pick the best-scoring
> topology from the 100 as the final ML tree.
> *How should I set up RAxML-NG?* In most cases I would like to use a
> single node but multiple cores. (Should MPI be used for this purpose?)

You will have to use a bot of scripting to invoke as many sequential
raxml-ng jobs as you have cores and collect the results in the end.

> And a theoretical question:
> *How many replicates one should infer?*

That really depends on the signal in your data, if most ML trees are
similar in terms of RF distance, then you are fine, if they are fairly
different, despite having similar likelihood scores (e.g., based on a
statistical significance test), then I'd do more tree searches.

> Attached are the files resulting from two analyses: a small one and a
> big one. For the big one I could attach a job statistics figure too,
> which shows that all 5 cores are working 100% all the time.

I had a quick look at them, with this type of short MSAs you should
really just use one core per raxml-ng run, i.e., if you have 16 cores
invoke 16 independent sequential raxml-ng runs.

>
> Best regards and thank you in advance,
> Gyozo
>
> Ps: Just read today's digest about the "number of free parameters" error
> message. Had the same error with the big one, that is why the --force
> mode is ON.

Alexey will fix that, but keep in mind that it's nothing else but a warning,

Alexis

>
> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology

www.exelixis-lab.org

Alexey Kozlov

unread,
Nov 30, 2018, 11:56:34 AM11/30/18
to ra...@googlegroups.com
Dear Gyozo,

just a small addition to Alexis' answers:

> Following your advice I moved (back) to RAxML from Exabayes (thank you
> for the detailed answer!) for the analysis of short alignments (few
> hundred residues long). The alignments might be short, but contain often
> numerous (thousands of) taxa. Thus I would like to use some
> parallelization. I did love that RAxML 8.X could infer the requested
> number of trees parallel on different CPU cores. You also suggest such
> parallelization in the manual, and I did try to achieve such an analysis
> setup. But I am not sure I succeeded, it seems, that RAxML-NG did use
> fine-grained parallelization and split my short alignment.
>
> So my aim is to set up RAxML-NG for parallel tree searching: if I need
> e.g. 100 trees, 1 core infers 50 trees and the other core the other 50
> trees parallel. At the end RAxML-NG should pick the best-scoring
> topology from the 100 as the final ML tree.
> *How should I set up RAxML-NG?*

Please have a look at the tutorial on coarse-grained parallelization
which I wrote just today :)

https://github.com/amkozlov/raxml-ng/wiki/Parallelization#coarse-grained-parallelization-for-short-alignments

A couple of comment regarding your analysis:

- I'd use GTR+G instead of JC since the number of additional free
parameters is negligible compared to the # branch lengths

- you should probably keep using 5 cores, or even try a bit more, since
it already takes too long

- dataset dimensions are very unfortunate not only for parallelization:
with almost 20,000 taxa and just 779 alignment sites, there is way too
little signal to resolve the tree, and thus we see poor/slow convergence

>In most cases I would like to use a
> single node but multiple cores. (Should MPI be used for this purpose?)

MPI version could be used on a single node, but I would use non-MPI
version in this case just to keep things simple.

Hope this helps,
Alexey

Győző Kaján

unread,
Dec 7, 2018, 11:15:46 AM12/7/18
to raxml
Dear Alex*,

Thank you very much for the detailed answer!
However, if I am honest, this feels like a major step back compared to RAxML 8. Why do I need to "hack" the software? Might be easy for you guys, but not for a noob/me. I mean I am sure I can solve it in 1 or 2 days, but RAxML 8 could do all this from one command prompt line: run parallel searches on independent cores and choose the best tree. Couldn't we get back this nice feature?

And a theoretical question again concerning the number of replicates: can you set a threshold in RAxML so it stops when the likelihood plateau/max is reached? Sg like the topological convergence diagnostics in ExaBayes. Otherwise it can be quite frustrating to determine the required number of generations. You set n: no, this was not enough. So you add another n. Oh, this was partially already a waste of wall-time...

Thank you very much again for your precious time and the detailed answers!!

Cheers,
Gyozo

Alexey Kozlov

unread,
Dec 7, 2018, 1:01:11 PM12/7/18
to ra...@googlegroups.com
Dear Gyozo,

thanks for your feedback.

Both features are on our TODO list. Please have understanding that we
cannot do everything at once. If you want an one-liner solution, please
have a look at our ParGenes pipeline:

https://github.com/BenoitMorel/ParGenes

Best,
Alexey

On 07.12.18 17:15, Győző Kaján wrote:
> Dear Alex*,
>
> Thank you very much for the detailed answer!
> However, if I am honest, this feels like a major step back compared to
> RAxML 8. Why do I need to "hack" the software? Might be easy for you
> guys, but not for a noob/me. I mean I am sure I can solve it in 1 or 2
> days, but RAxML 8 could do all this from one command prompt line: run
> parallel searches on independent cores and choose the best tree.
> Couldn't we get back this nice feature?
>
> And a theoretical question again concerning the number of replicates:
> can you set a threshold in RAxML so it stops when the likelihood
> plateau/max is reached? Sg like the topological convergence
> diagnostics in ExaBayes. Otherwise it can be quite frustrating to
> determine the required number of generations. You set /n/: no, this was
> not enough. So you add another /n/. Oh, this was partially already a

Alexandros Stamatakis

unread,
Dec 9, 2018, 6:37:53 AM12/9/18
to ra...@googlegroups.com
Hi Győző,

> Thank you very much for the detailed answer!
> However, if I am honest, this feels like a major step back compared to
> RAxML 8. Why do I need to "hack" the software? Might be easy for you
> guys, but not for a noob/me. I mean I am sure I can solve it in 1 or 2
> days, but RAxML 8 could do all this from one command prompt line: run
> parallel searches on independent cores and choose the best tree.
> Couldn't we get back this nice feature?

We will get back to this at some point, RAxML urgently needed a
re-design though, and we can't just re-implement all the features of a
code developed over 15 years that quickly.

> And a theoretical question again concerning the number of replicates:
> can you set a threshold in RAxML so it stops when the likelihood
> plateau/max is reached? Sg like the topological convergence
> diagnostics in ExaBayes. Otherwise it can be quite frustrating to
> determine the required number of generations. You set /n/: no, this was
> not enough. So you add another /n/. Oh, this was partially already a
> waste of wall-time...

This is not that easy to implement in a ML framework, but we will think
about it.

Alexis
> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

Kaján Győző

unread,
Dec 10, 2018, 4:59:41 PM12/10/18
to ra...@googlegroups.com
Dear Alex*,

Sorry, I didn't want to be impatient. It already half satisfies me, that it is not only me, missing these features!
Thank you, and keep up the good work!

Cheers,
Gyozo

Mentes a vírusoktól. www.avg.com

You received this message because you are subscribed to a topic in the Google Groups "raxml" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

Mentes a vírusoktól. www.avg.com

Kaján Győző

unread,
Dec 28, 2018, 7:41:49 PM12/28/18
to ra...@googlegroups.com
Dear Alexey,

Hope you have nice holidays, and sorry to bother you!
But I tried the recommended setup, unfortunately it did not work for me. Seemingly it ran the tree searches sequentially and not parallelly. Could you pls check my files, what did I do wrong?
My hardware:
Snowy provides 3648 cores in the form of 228 dual CPU (Intel Xeon E5-2660) HP ProLiant SL230s Gen8 nodes with a default memory configuration of 128 Gigabyte. Each CPU has 8 cores.
Happy New Year!
 
Thx,
Gyozo


bash_HAdV_srun16.txt
slurm-140103.out

Alexey Kozlov

unread,
Jan 7, 2019, 9:27:42 AM1/7/19
to ra...@googlegroups.com
Dear Gyozo,

please try to change submission script header as follows:

#SBATCH -n 16
#SBATCH --cpus-per-task=1

(i.e. 16 tasks x 1 core instead of 1 task x 16 cores).

Interestingly, both configurations work fine on our cluster, but this
might depend on SLURM version and/or configuration (we have
slurm/17.11.7.1 here).

Please let me know if the alternative solution above works for you - if
yes, I'll change the tutorial accordingly.

Thanks in advance!

Best,
Alexey

On 29.12.18 01:41, Kaján Győző wrote:
> Dear Alexey,
>
> Hope you have nice holidays, and sorry to bother you!
> But I tried the recommended
> <https://github.com/amkozlov/raxml-ng/wiki/Parallelization#coarse-grained-parallelization-for-short-alignments>
> setup, unfortunately it did not work for me. Seemingly it ran the tree
> searches sequentially and not parallelly. Could you pls check my files,
> what did I do wrong?
> My hardware:
> Snowy
> <https://www.uppmax.uu.se/resources/systems/the-snowy-cluster/>provides
> 3648 cores in the form of 228 dual CPU (Intel Xeon E5-2660) HP ProLiant
> SL230s Gen8 nodes with a default memory configuration of 128 Gigabyte.
> Each CPU has 8 cores.
> Happy New Year!
>
> Thx,
> Gyozo
>
>
>
> On Mon, 10 Dec 2018 at 22:59, Kaján Győző <gyk...@gmail.com
> <mailto:gyk...@gmail.com>> wrote:
>
> Dear Alex*,
>
> Sorry, I didn't want to be impatient. It already half satisfies me,
> that it is not only me, missing these features!
> Thank you, and keep up the good work!
>
> Cheers,
> Gyozo
>
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> Mentes a vírusoktól. www.avg.com
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
>
>
> <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
> On Sun, 9 Dec 2018 at 12:37, Alexandros Stamatakis
> <alexandros...@gmail.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> > <mailto:raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>>.
> > For more options, visit https://groups.google.com/d/optout.
>
> --
> Alexandros (Alexis) Stamatakis
>
> Research Group Leader, Heidelberg Institute for Theoretical Studies
> Full Professor, Dept. of Informatics, Karlsruhe Institute of
> Technology
>
> www.exelixis-lab.org <http://www.exelixis-lab.org>
>
> --
> You received this message because you are subscribed to a topic
> in the Google Groups "raxml" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
> To unsubscribe from this group and all its topics, send an email
> to raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.
>
>
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> Mentes a vírusoktól. www.avg.com
> <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
>
>
> <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>

Kaján Győző

unread,
Jan 14, 2019, 11:14:50 AM1/14/19
to ra...@googlegroups.com
Dear Alexey,

Tried it, but did no help. Strange, because it seems, it starts the trees in parallel but on a single core. See job statistics attached for the prev. run too.
We have slurm 17.11.12.

Thx,
Gyozo


To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.
bash_HAdV_srun16.txt
snowy-snic2017-7-285-gyozo-144167.png
slurm-144167.out
snowy-snic2017-7-285-gyozo-140103.png

Alexey Kozlov

unread,
Jan 14, 2019, 12:48:03 PM1/14/19
to ra...@googlegroups.com
Hi Gyozo,

very weird, apparently there are some differences in configuration.

Could you please try

srun -N 1 -n 1 --exclusive raxml-ng ...

and/ or

srun -N 1 -n 1 --cpu-bind=cores raxml-ng ...

According to SLURM documentation, this should help:

https://slurm.schedmd.com/srun.html#OPT_cpu-bind

Best,
Alexey

On 14.01.19 17:14, Kaján Győző wrote:
> Dear Alexey,
>
> Tried it, but did no help. Strange, because it seems, it starts the
> trees in parallel but on a single core. See job statistics attached for
> the prev. run too.
> We have slurm 17.11.12.
>
> Thx,
> Gyozo
>
>
> On Mon, 7 Jan 2019 at 15:27, Alexey Kozlov <alexei...@gmail.com
> <mailto:alexei...@gmail.com>> wrote:
>
> Dear Gyozo,
>
> please try to change submission script header as follows:
>
> #SBATCH -n 16
> #SBATCH --cpus-per-task=1
>
> (i.e. 16 tasks x 1 core instead of 1 task x 16 cores).
>
> Interestingly, both configurations work fine on our cluster, but this
> might depend on SLURM version and/or configuration (we have
> slurm/17.11.7.1 <http://17.11.7.1> here).
> > <mailto:gyk...@gmail.com <mailto:gyk...@gmail.com>>> wrote:
> >
> >     Dear Alex*,
> >
> >     Sorry, I didn't want to be impatient. It already half
> satisfies me,
> >     that it is not only me, missing these features!
> >     Thank you, and keep up the good work!
> >
> >     Cheers,
> >     Gyozo
> >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >       Mentes a vírusoktól. www.avg.com <http://www.avg.com>
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >
> >
> >     <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >
> >     On Sun, 9 Dec 2018 at 12:37, Alexandros Stamatakis
> >     <alexandros...@gmail.com
> <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com
> >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >          > <mailto:raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>>.
> >          > For more options, visit
> https://groups.google.com/d/optout.
> >
> >         --
> >         Alexandros (Alexis) Stamatakis
> >
> >         Research Group Leader, Heidelberg Institute for
> Theoretical Studies
> >         Full Professor, Dept. of Informatics, Karlsruhe Institute of
> >         Technology
> >
> > www.exelixis-lab.org <http://www.exelixis-lab.org>
> <http://www.exelixis-lab.org>
> >
> >         --
> >         You received this message because you are subscribed to a
> topic
> >         in the Google Groups "raxml" group.
> >         To unsubscribe from this topic, visit
> > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
> >         To unsubscribe from this group and all its topics, send
> an email
> >         to raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>.
> >         For more options, visit https://groups.google.com/d/optout.
> >
> >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >       Mentes a vírusoktól. www.avg.com <http://www.avg.com>
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >
> >
> >     <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >
> > --
> > You received this message because you are subscribed to the Google
> > Groups "raxml" group.
> > To unsubscribe from this group and stop receiving emails from it,
> send
> > an email to raxml+un...@googlegroups.com
> You received this message because you are subscribed to a topic in
> the Google Groups "raxml" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.
>

Kaján Győző

unread,
Jan 14, 2019, 5:08:13 PM1/14/19
to ra...@googlegroups.com
Dear Alexey,

The first was the winner one:
srun -N 1 -n 1 --exclusive raxml-ng  ...  

Complete runtime: 58 sec.

Thank you very much!
Gyozo

To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.
slurm-146119.txt

Alexey Kozlov

unread,
Jan 14, 2019, 8:12:25 PM1/14/19
to ra...@googlegroups.com
Great, thank you for testing this, I will update the tutorial accordingly!

Alexey

On 14.01.19 23:07, Kaján Győző wrote:
> Dear Alexey,
>
> The first was the winner one:
> srun -N 1 -n 1 --*exclusive *raxml-ng  ...
> > <mailto:alexei...@gmail.com
> <mailto:alexei...@gmail.com>>> wrote:
> >
> >     Dear Gyozo,
> >
> >     please try to change submission script header as follows:
> >
> >     #SBATCH -n 16
> >     #SBATCH --cpus-per-task=1
> >
> >     (i.e. 16 tasks x 1 core instead of 1 task x 16 cores).
> >
> >     Interestingly, both configurations work fine on our cluster,
> but this
> >     might depend on SLURM version and/or configuration (we have
> >     slurm/17.11.7.1 <http://17.11.7.1> <http://17.11.7.1> here).
> <http://www.avg.com> <http://www.avg.com>
> >      >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>
> >      >          > <mailto:raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >      >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>>.
> >      >         <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>.
> >      >         For more options, visit
> https://groups.google.com/d/optout.
> >      >
> >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >      >       Mentes a vírusoktól. www.avg.com
> <http://www.avg.com> <http://www.avg.com>
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >      >
> >      >
> >      >
>  <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >      >
> >      > --
> >      > You received this message because you are subscribed to
> the Google
> >      > Groups "raxml" group.
> >      > To unsubscribe from this group and stop receiving emails
> from it,
> >     send
> >      > an email to raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >      > <mailto:raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>>.
> >      > For more options, visit https://groups.google.com/d/optout.
> >
> >     --
> >     You received this message because you are subscribed to a
> topic in
> >     the Google Groups "raxml" group.
> >     To unsubscribe from this topic, visit
> > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
> >     To unsubscribe from this group and all its topics, send an
> email to
> > raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>.
> >     For more options, visit https://groups.google.com/d/optout.
> >

Kaján Győző

unread,
Feb 8, 2019, 7:24:14 AM2/8/19
to ra...@googlegroups.com
Dear Alexey,

How could I go beyond one node? I have tried -N 2 or -n 32, but not all cores were used.

Thx,
Gyozo

To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.

Alexey Kozlov

unread,
Feb 12, 2019, 8:08:40 AM2/12/19
to ra...@googlegroups.com
Hi Gyozo,

for multiple nodes, I recommend using our ParGenes pipeline:

https://github.com/BenoitMorel/ParGenes

Or please feel free to dig into SLURM documentation and share your
findings here :)

Best,
Alexey

On 07.02.19 22:43, Kaján Győző wrote:
> Dear Alexey,
>
> How could I go beyond one node? I have tried -N 2 or -n 32, but not all
> cores were used.
>
> Thx,
> Gyozo
>
> On Tue, 15 Jan 2019 at 02:12, Alexey Kozlov <alexei...@gmail.com
> <mailto:alexei...@gmail.com>> wrote:
>
> Great, thank you for testing this, I will update the tutorial
> accordingly!
>
> Alexey
>
> On 14.01.19 23:07, Kaján Győző wrote:
> > Dear Alexey,
> >
> > The first was the winner one:
> > srun -N 1 -n 1 --*exclusive *raxml-ng  ...
> >
> > Complete runtime: 58 sec.
> >
> > Thank you very much!
> > Gyozo
> >
> > On Mon, 14 Jan 2019 at 18:48, Alexey Kozlov
> <alexei...@gmail.com <mailto:alexei...@gmail.com>
> >      >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>
> >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>>>>
> >      >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>
> >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>>>>>.
> >      >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>
> >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>>>>.
> >      >      >         For more options, visit
> > https://groups.google.com/d/optout.
> >      >      >
> >      >      >
> >      >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >      >      >       Mentes a vírusoktól. www.avg.com
> <http://www.avg.com>
> >     <http://www.avg.com> <http://www.avg.com>
> >      >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
> >      >      >
> >      >      >
> >      >      >
> >       <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >      >      >
> >      >      > --
> >      >      > You received this message because you are subscribed to
> >     the Google
> >      >      > Groups "raxml" group.
> >      >      > To unsubscribe from this group and stop receiving
> emails
> >     from it,
> >      >     send
> >      >      > an email to raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >      >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>
> >      >      > <mailto:raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >      >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>>.
> >      >      > For more options, visit
> https://groups.google.com/d/optout.
> >      >
> >      >     --
> >      >     You received this message because you are subscribed to a
> >     topic in
> >      >     the Google Groups "raxml" group.
> >      >     To unsubscribe from this topic, visit
> >      >
> https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
> >      >     To unsubscribe from this group and all its topics, send an
> >     email to
> >      > raxml+un...@googlegroups.com
> <mailto:raxml%2Bunsu...@googlegroups.com>
> >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>>
> >      >     <mailto:raxml%2Bunsu...@googlegroups.com
> <mailto:raxml%252Buns...@googlegroups.com>
> >     <mailto:raxml%252Buns...@googlegroups.com
> <mailto:raxml%25252Bun...@googlegroups.com>>>.
> >      >     For more options, visit
> https://groups.google.com/d/optout.
> >      >

Kaján Győző

unread,
May 27, 2019, 7:40:41 AM5/27/19
to ra...@googlegroups.com
Dear Alexey,

Running the attached bash file on the same dataset
  • in January, the calculation was running on 20 cores parallelly.
  • now, it runs the 20 "units" of work sequentially, one after the other.
What is your gut feeling, what system change might cause this strange behaviour?

We have Slurm 17.11.13 now, no idea about the version No of January.

Best regards,
Gyozo

To unsubscribe from this group and all its topics, send an email to raxml+un...@googlegroups.com.
bash_RAxML_NG_ORF5.txt

Győző Kaján

unread,
May 28, 2019, 9:08:25 AM5/28/19
to raxml
Dear Alexey,

Found the solution. Are you sure this srun workaround is needed? It is working wonderfully for me without it (see below). And you don't have srun in the bootstrap part either. And found the solution for the multiple nodes as well: the -p does the trick at least on our Slurm (v17.11.13) system. In the example below, the tree inferences are running on 2 nodes (40 cores altogether) parallelly.

Best regards,
Gyozo


#! /bin/bash -l
#SBATCH -A XXXXXXX
#SBATCH -J PRRSV_ORF5
#SBATCH -t 10:00:00
#SBATCH --mail-type=ALL
#SBATCH -p node
#SBATCH -n 40
#SBATCH -B 2:10:1
#SBATCH --threads-per-core=1
#SBATCH --cpus-per-task=1
#SBATCH --mail-type=ALL
#SBATCH --qos=short
#SBATCH -M snowy


module load bioinfo-tools
module load RAxML-NG/0.7.0b


cd /domus/h1/gyozo/Feri/PRRSV1_ORF5/


for i in `seq 1 40`;
do
 raxml-ng --msa alignment.phy --model GTR+G+I --force --prefix parallel_$i --threads 1 --seed $RANDOM &
done

wait




>     <mailto:alexei.kozlow@gmail.com>>> wrote:
>      >
>      >     Hi Gyozo,
>      >
>      >     very weird, apparently there are some differences in
>     configuration.
>      >
>      >     Could you please try
>      >
>      >     srun -N 1 -n 1 --exclusive raxml-ng  ...
>      >
>      >     and/ or
>      >
>      >     srun -N 1 -n 1 --cpu-bind=cores raxml-ng  ...
>      >
>      >     According to SLURM documentation, this should help:
>      >
>      > https://slurm.schedmd.com/srun.html#OPT_cpu-bind
>      >
>      >     Best,
>      >     Alexey
>      >
>      >     On 14.01.19 17:14, Kaján Győző wrote:
>      >      > Dear Alexey,
>      >      >
>      >      > Tried it, but did no help. Strange, because it seems, it
>     starts the
>      >      > trees in parallel but on a single core. See job statistics
>      >     attached for
>      >      > the prev. run too.
>      >      > We have slurm 17.11.12.
>      >      >
>      >      > Thx,
>      >      > Gyozo
>      >      >
>      >      >
>      >      > On Mon, 7 Jan 2019 at 15:27, Alexey Kozlov
>      >     <alexei...@gmail.com <mailto:alexei.kozlow@gmail.com>
>     <mailto:alexei.kozlow@gmail.com <mailto:alexei.kozlow@gmail.com>>
>      >      > <mailto:alexei.kozlow@gmail.com
>     <mailto:alexei.kozlow@gmail.com>
>      >     <mailto:alexei.kozlow@gmail.com
>      >      >      >     <alexandros.stamatakis@gmail.com
>     <mailto:alexandros.stama...@gmail.com>
>      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>>
>      >      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>
>      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>>>
>      >      >      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>
>      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>>
>      >      >     <mailto:alexandros.stama...@gmail.com
>     <mailto:alexandros.stama...@gmail.com>
>      >     <mailto:alexandros.stama...@gmail.com
>     raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>>
>      >      >      >       
>       <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%25252Bunsubscribe@googlegroups.com
>     <mailto:raxml%2525252Bunsubscribe@googlegroups.com>>>>
>      >      >      >          >
>     <mailto:raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>>
>      >      >      >       
>       <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%25252Bunsubscribe@googlegroups.com
>     <mailto:raxml%2525252Bunsubscribe@googlegroups.com>>>>>.

>      >      >      >          > For more options, visit
>      >      > https://groups.google.com/d/optout.
>      >      >      >
>      >      >      >         --
>      >      >      >         Alexandros (Alexis) Stamatakis
>      >      >      >
>      >      >      >         Research Group Leader, Heidelberg Institute for
>      >      >     Theoretical Studies
>      >      >      >         Full Professor, Dept. of Informatics, Karlsruhe
>      >     Institute of
>      >      >      >         Technology
>      >      >      >
>      >      >      > www.exelixis-lab.org <http://www.exelixis-lab.org>
>     <http://www.exelixis-lab.org>
>      >     <http://www.exelixis-lab.org>
>      >      >     <http://www.exelixis-lab.org>
>      >      >      >
>      >      >      >         --
>      >      >      >         You received this message because you are
>      >     subscribed to a
>      >      >     topic
>      >      >      >         in the Google Groups "raxml" group.
>      >      >      >         To unsubscribe from this topic, visit
>      >      >      >
>      > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
>      >      >      >         To unsubscribe from this group and all its
>     topics,
>      >     send
>      >      >     an email
>      >      >      >         to raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>>
>      >      >      >       
>       <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%25252Bunsubscribe@googlegroups.com
>     <mailto:raxml%2525252Bunsubscribe@googlegroups.com>>>>.

>      >      >      >         For more options, visit
>      > https://groups.google.com/d/optout.
>      >      >      >
>      >      >      >
>      >      >      >
>      >      >
>      >     
>       <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
>      >      >      >       Mentes a vírusoktól. www.avg.com
>     <http://www.avg.com>
>      >     <http://www.avg.com> <http://www.avg.com>
>      >      >      >
>      >      >
>      >     
>       <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
>      >      >      >
>      >      >      >
>      >      >      >
>      >       <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>      >      >      >
>      >      >      > --
>      >      >      > You received this message because you are subscribed to
>      >     the Google
>      >      >      > Groups "raxml" group.
>      >      >      > To unsubscribe from this group and stop receiving
>     emails
>      >     from it,
>      >      >     send
>      >      >      > an email to raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>>
>      >      >      > <mailto:raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%252Bunsubscribe@googlegroups.com
>     <mailto:raxml%25252Bunsubscribe@googlegroups.com>>>>.

>      >      >      > For more options, visit
>     https://groups.google.com/d/optout.
>      >      >
>      >      >     --
>      >      >     You received this message because you are subscribed to a
>      >     topic in
>      >      >     the Google Groups "raxml" group.
>      >      >     To unsubscribe from this topic, visit
>      >      >
>     https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
>      >      >     To unsubscribe from this group and all its topics, send an
>      >     email to

>      >      >     For more options, visit
>     https://groups.google.com/d/optout.
>      >      >
>      >      > --
>      >      > You received this message because you are subscribed to
>     the Google
>      >      > Groups "raxml" group.
>      >      > To unsubscribe from this group and stop receiving emails
>     from it,
>      >     send
>      >      > an email to raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>
>      >      > <mailto:raxml+unsubscribe@googlegroups.com
>     <mailto:raxml%2Bunsubscribe@googlegroups.com>
>      >     <mailto:raxml%2Bunsubscribe@googlegroups.com
>     <mailto:raxml%252Bunsubscribe@googlegroups.com>>>.

>      >      > For more options, visit https://groups.google.com/d/optout.
>      >
>      >     --
>      >     You received this message because you are subscribed to a
>     topic in
>      >     the Google Groups "raxml" group.
>      >     To unsubscribe from this topic, visit
>      > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
>      >     To unsubscribe from this group and all its topics, send an
>     email to

>      >     For more options, visit https://groups.google.com/d/optout.
>      >
>      > --
>      > You received this message because you are subscribed to the Google
>      > Groups "raxml" group.
>      > To unsubscribe from this group and stop receiving emails from it,
>     send

>      > For more options, visit https://groups.google.com/d/optout.
>
>     --
>     You received this message because you are subscribed to a topic in
>     the Google Groups "raxml" group.
>     To unsubscribe from this topic, visit
>     https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
>     To unsubscribe from this group and all its topics, send an email to

>     For more options, visit https://groups.google.com/d/optout.
>
> --
> You received this message because you are subscribed to the Google
> Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send

> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "raxml" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to raxml+unsubscribe@googlegroups.com.

Alexey Kozlov

unread,
May 28, 2019, 10:18:00 AM5/28/19
to ra...@googlegroups.com
Dear Gyozo,

many thanks for sharing your solution!

I just double-checked, and on our system (SLURM 18.08.4, CentOS 7.6.1810) both options - with and
without "srun -N 1 -n 1" - work just fine.

In any case, I'm currently working on implementing coarse-grained parallelization directly in
raxml-ng, so I hope these problems will become irrelevant soon :)

Best,
Alexey

On 28.05.19 15:08, Győző Kaján wrote:
> Dear Alexey,
>
> Found the solution. Are you sure this srun workaround is needed? It is working wonderfully for me
> without it (see below). And you don't have srun in the bootstrap
> <https://github.com/amkozlov/raxml-ng/wiki/Parallelization> part either. And found the solution for
> the multiple nodes as well: the -p does the trick at least on our Slurm (v17.11.13) system. In the
> example below, the tree inferences are running on 2 nodes (40 cores altogether) parallelly.
>
> Best regards,
> Gyozo
>
>
> #! /bin/bash -l
> #SBATCH -A XXXXXXX
> #SBATCH -J PRRSV_ORF5
> #SBATCH -t 10:00:00
> #SBATCH --mail-type=ALL
> #SBATCH *-p node*
> #SBATCH -n 40
> #SBATCH -B 2:10:1
> #SBATCH --threads-per-core=1
> #SBATCH --cpus-per-task=1
> #SBATCH --mail-type=ALL
> #SBATCH --qos=short
> #SBATCH -M snowy
>
>
> module load bioinfo-tools
> module load RAxML-NG/0.7.0b
>
>
> cd /domus/h1/gyozo/Feri/PRRSV1_ORF5/
>
>
> for i in `seq 1 40`;
> do
>  raxml-ng --msa alignment.phy --model GTR+G+I --force --prefix parallel_$i --threads 1 --seed $RANDOM &
> done
>
> wait
>
>
>
>
>
> On Monday, 27 May 2019 13:40:41 UTC+2, Győző Kaján wrote:
>
> Dear Alexey,
>
> Running the attached bash file on the same dataset
>
> * in January, the calculation was running on 20 cores *parallelly*.
> * now, it runs the 20 "units" of work *sequentially*, one after the other.
>
> What is your gut feeling, what system change might cause this strange behaviour?
>
> We have Slurm 17.11.13 now, no idea about the version No of January.
>
> Best regards,
> Gyozo
>
> On Tue, 12 Feb 2019 at 14:08, Alexey Kozlov <alexei...@gmail.com
> <mailto:alexei...@gmail.com>> wrote:
>
> Hi Gyozo,
>
> for multiple nodes, I recommend using our ParGenes pipeline:
>
> https://github.com/BenoitMorel/ParGenes <https://github.com/BenoitMorel/ParGenes>
>
> Or please feel free to dig into SLURM documentation and share your
> findings here :)
>
> Best,
> Alexey
>
> On 07.02.19 22:43, Kaján Győző wrote:
> > Dear Alexey,
> >
> > How could I go beyond one node? I have tried -N 2 or -n 32, but not all
> > cores were used.
> >
> > Thx,
> > Gyozo
> >
> > On Tue, 15 Jan 2019 at 02:12, Alexey Kozlov <alexei...@gmail.com
> <mailto:alexei...@gmail.com>
> > <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>>> wrote:
> >
> >     Great, thank you for testing this, I will update the tutorial
> >     accordingly!
> >
> >     Alexey
> >
> >     On 14.01.19 23:07, Kaján Győző wrote:
> >      > Dear Alexey,
> >      >
> >      > The first was the winner one:
> >      > srun -N 1 -n 1 --*exclusive *raxml-ng  ...
> >      >
> >      > Complete runtime: 58 sec.
> >      >
> >      > Thank you very much!
> >      > Gyozo
> >      >
> >      > On Mon, 14 Jan 2019 at 18:48, Alexey Kozlov
> >     <alexei...@gmail.com <mailto:alexei...@gmail.com>
> <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>>
> >      > <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>
> >     <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>>>> wrote:
> >      >
> >      >     Hi Gyozo,
> >      >
> >      >     very weird, apparently there are some differences in
> >     configuration.
> >      >
> >      >     Could you please try
> >      >
> >      >     srun -N 1 -n 1 --exclusive raxml-ng  ...
> >      >
> >      >     and/ or
> >      >
> >      >     srun -N 1 -n 1 --cpu-bind=cores raxml-ng  ...
> >      >
> >      >     According to SLURM documentation, this should help:
> >      >
> >      > https://slurm.schedmd.com/srun.html#OPT_cpu-bind
> <https://slurm.schedmd.com/srun.html#OPT_cpu-bind>
> >      >
> >      >     Best,
> >      >     Alexey
> >      >
> >      >     On 14.01.19 17:14, Kaján Győző wrote:
> >      >      > Dear Alexey,
> >      >      >
> >      >      > Tried it, but did no help. Strange, because it seems, it
> >     starts the
> >      >      > trees in parallel but on a single core. See job statistics
> >      >     attached for
> >      >      > the prev. run too.
> >      >      > We have slurm 17.11.12.
> >      >      >
> >      >      > Thx,
> >      >      > Gyozo
> >      >      >
> >      >      >
> >      >      > On Mon, 7 Jan 2019 at 15:27, Alexey Kozlov
> >      >     <alexei...@gmail.com <mailto:alexei...@gmail.com>
> >      >      > <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>
> >     <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>>
> >      >     <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>
> >     <mailto:alexei...@gmail.com <mailto:alexei...@gmail.com>>>>> wrote:
> >      >      >
> >      >      >     Dear Gyozo,
> >      >      >
> >      >      >     please try to change submission script header as follows:
> >      >      >
> >      >      >     #SBATCH -n 16
> >      >      >     #SBATCH --cpus-per-task=1
> >      >      >
> >      >      >     (i.e. 16 tasks x 1 core instead of 1 task x 16 cores).
> >      >      >
> >      >      >     Interestingly, both configurations work fine on our
> >     cluster,
> >      >     but this
> >      >      >     might depend on SLURM version and/or configuration (we
> >     have
> >      >      >     slurm/17.11.7.1 <http://17.11.7.1> <http://17.11.7.1> <http://17.11.7.1>
> >     <http://17.11.7.1> here).
> >      >      >
> >      >      >     Please let me know if the alternative solution above works
> >      >     for you - if
> >      >      >     yes, I'll change the tutorial accordingly.
> >      >      >
> >      >      >     Thanks in advance!
> >      >      >
> >      >      >     Best,
> >      >      >     Alexey
> >      >      >
> >      >      >     On 29.12.18 01:41, Kaján Győző wrote:
> >      >      >      > Dear Alexey,
> >      >      >      >
> >      >      >      > Hope you have nice holidays, and sorry to bother you!
> >      >      >      > But I tried the recommended
> >      >      >      >
> >      >      >
> >      >
> >
>  <https://github.com/amkozlov/raxml-ng/wiki/Parallelization#coarse-grained-parallelization-for-short-alignments <https://github.com/amkozlov/raxml-ng/wiki/Parallelization#coarse-grained-parallelization-for-short-alignments>>
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>>
> >      >      >      >       Mentes a vírusoktól. www.avg.com <http://www.avg.com>
> >     <http://www.avg.com>
> >      >     <http://www.avg.com> <http://www.avg.com>
> >      >      >      >
> >      >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>>
> >      >      >      >
> >      >      >      >
> >      >      >      >
> >      >       <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >      >      >      >
> >      >      >      >     On Sun, 9 Dec 2018 at 12:37, Alexandros Stamatakis
> >      >      >      >     <alexandros...@gmail.com
> <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>
> >      >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>>
> >      >      >     <mailto:alexandros...@gmail.com
> <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>
> >      >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>>>
> >      >      >      >     <mailto:alexandros...@gmail.com
> <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>
> >      >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>>
> >      >      >     <mailto:alexandros...@gmail.com
> <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>
> >      >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>
> >     <mailto:alexandros...@gmail.com <mailto:alexandros...@gmail.com>>>>>>
> >      >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>
> >     <mailto:raxml%2525252Bu...@googlegroups.com
> <mailto:raxml%252525252B...@googlegroups.com>>>>>
> >      >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>
> >     <mailto:raxml%2525252Bu...@googlegroups.com
> <mailto:raxml%252525252B...@googlegroups.com>>>>>>.
> >      >      >      >          > For more options, visit
> >      >      > https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
> >      >      >      >
> >      >      >      >         --
> >      >      >      >         Alexandros (Alexis) Stamatakis
> >      >      >      >
> >      >      >      >         Research Group Leader, Heidelberg Institute for
> >      >      >     Theoretical Studies
> >      >      >      >         Full Professor, Dept. of Informatics, Karlsruhe
> >      >     Institute of
> >      >      >      >         Technology
> >      >      >      >
> >      >      >      > www.exelixis-lab.org <http://www.exelixis-lab.org>
> <http://www.exelixis-lab.org>
> >     <http://www.exelixis-lab.org>
> >      >     <http://www.exelixis-lab.org>
> >      >      >     <http://www.exelixis-lab.org>
> >      >      >      >
> >      >      >      >         --
> >      >      >      >         You received this message because you are
> >      >     subscribed to a
> >      >      >     topic
> >      >      >      >         in the Google Groups "raxml" group.
> >      >      >      >         To unsubscribe from this topic, visit
> >      >      >      >
> >      > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe
> <https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe>.
> >      >      >      >         To unsubscribe from this group and all its
> >     topics,
> >      >     send
> >      >      >     an email
> >      >      >      >         to raxml+un...@googlegroups.com
> >      >     <mailto:raxml%25252Bun...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>
> >     <mailto:raxml%2525252Bu...@googlegroups.com
> <mailto:raxml%252525252B...@googlegroups.com>>>>>.
> >      >      >      >         For more options, visit
> >      > https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
> >      >      >      >
> >      >      >      >
> >      >      >      >
> >      >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>>
> >      >      >      >       Mentes a vírusoktól. www.avg.com <http://www.avg.com>
> >     <http://www.avg.com>
> >      >     <http://www.avg.com> <http://www.avg.com>
> >      >      >      >
> >      >      >
> >      >
> >
>  <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>>
> >      >      >      >
> >      >      >      >
> >      >      >      >
> >      >       <#m_8355361735492220274_DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> >      >      >      >
> >      >      >      > --
> >      >      >      > You received this message because you are subscribed to
> >      >     the Google
> >      >      >      > Groups "raxml" group.
> >      >      >      > To unsubscribe from this group and stop receiving
> >     emails
> >      >     from it,
> >      >      >     send
> >      >      >      > an email to raxml+un...@googlegroups.com
> <mailto:raxml%2525252Bu...@googlegroups.com>>>>>.
> >      >      >      > For more options, visit
> > https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
> >      >      >
> >      >      >     --
> >      >      >     You received this message because you are subscribed to a
> >      >     topic in
> >      >      >     the Google Groups "raxml" group.
> >      >      >     To unsubscribe from this topic, visit
> >      >      >
> > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe
> <https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe>.
> >      >      >     To unsubscribe from this group and all its topics, send an
> >      >     email to
> <mailto:raxml%2525252Bu...@googlegroups.com>>>>.
> >      >      >     For more options, visit
> > https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
> >      >      >
> >      >      > --
> >      >      > You received this message because you are subscribed to
> >     the Google
> >      >      > Groups "raxml" group.
> >      >      > To unsubscribe from this group and stop receiving emails
> >     from it,
> >      >     send
> <mailto:raxml%25252Bun...@googlegroups.com>>>>.
> >      >      > For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
> >      >
> >      >     --
> >      >     You received this message because you are subscribed to a
> >     topic in
> >      >     the Google Groups "raxml" group.
> >      >     To unsubscribe from this topic, visit
> >      > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe
> <https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe>.
> >      >     To unsubscribe from this group and all its topics, send an
> >     email to
> <mailto:raxml%25252Bun...@googlegroups.com>>>.
> >      >     For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
> >      >
> >      > --
> >      > You received this message because you are subscribed to the Google
> >      > Groups "raxml" group.
> >      > To unsubscribe from this group and stop receiving emails from it,
> >     send
> <mailto:raxml%252Buns...@googlegroups.com>>>.
> >      > For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
> >
> >     --
> >     You received this message because you are subscribed to a topic in
> >     the Google Groups "raxml" group.
> >     To unsubscribe from this topic, visit
> > https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe
> <https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe>.
> >     To unsubscribe from this group and all its topics, send an email to
> <mailto:raxml%252Buns...@googlegroups.com>>.
> >     For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
> >
> > --
> > You received this message because you are subscribed to the Google
> > Groups "raxml" group.
> > To unsubscribe from this group and stop receiving emails from it, send
> > <mailto:raxml+un...@googlegroups.com <mailto:raxml%2Bunsu...@googlegroups.com>>.
> > For more options, visit https://groups.google.com/d/optout
> <https://groups.google.com/d/optout>.
>
> --
> You received this message because you are subscribed to a topic in the Google Groups "raxml"
> group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe
> <https://groups.google.com/d/topic/raxml/PYCTUdWJJOA/unsubscribe>.
> To unsubscribe from this group and all its topics, send an email to
> raxml+un...@googlegroups.com <mailto:raxml%2Bunsu...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout <https://groups.google.com/d/optout>.
>
> --
> You received this message because you are subscribed to the Google Groups "raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
> raxml+un...@googlegroups.com <mailto:raxml+un...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/raxml/e5fa4040-035b-4059-9289-20bcfa59a626%40googlegroups.com
> <https://groups.google.com/d/msgid/raxml/e5fa4040-035b-4059-9289-20bcfa59a626%40googlegroups.com?utm_medium=email&utm_source=footer>.
Reply all
Reply to author
Forward
0 new messages