compile error: error while loading shared libraries: libgmp.so.3

Hi,

Usually, you are not allowed to compile on a node.
So therefore, those path might not be available on the node.

Actually, madgraph (actually madevent) are able to dealt with the cluster themselves, and they know which part need to be run locally and which part can be run in parallel on the cluster.
You just need to configure the file
Cards/me5_configuration.txt
(for more permanent use, you can modify the one in the MG5 directory: input/mg5_configuration.txt)
and configure the following three variables: run_mode, cluster_type,cluster_queue as following

#! Default Running mode
#! 0: single machine/ 1: cluster / 2: multicore
 run_mode = 1

#! Cluster Type [pbs|sge|condor|lsf|ge|slurm] Use for cluster run only
#! And cluster queue
cluster_type = pbs
cluster_queue = ?????

Don't forget to remove the "#" at the beginning of the line.

Cheers,

Olivier

On Nov 18, 2013, at 5:46 PM, Kechen Wang <email address hidden> wrote:

> New question #239418 on MadGraph5:
> https://answers.launchpad.net/madgraph5/+question/239418
>
>
> When I qsub the following shell code to the cluster,
> " cd /home/kechen/MadGraph5_v1_5_13/pp2ttbar_parton_pgs_1/
> /home/kechen/Python-2.6/python ./bin/generate_events --laststep=pgs -f ",
>
> I got the following error message:
> ###################################
> Error detected in "generate_events --laststep=pgs -f"
> write debug file /home/kechen/MadGraph5_v1_5_13/pp2ttbar_parton_pgs_1/run_01_tag_1_debug.log
> If you need help with this issue please contact us on https://answers.launchpad.net/madgraph5
> MadGraph5Error : A compilation Error occurs when trying to compile /home/kechen/MadGraph5_v1_5_13/pp2ttbar_parton_pgs_1/Source.
> The compilation fails with the following output message:
> gfortran -O -w -ffixed-line-length-132 -c -o gen_ximprove.o gen_ximprove.f
> gfortran -O -w -ffixed-line-length-132 -c -o ranmar.o ranmar.f
> gfortran -O -w -ffixed-line-length-132 -c -o rw_routines.o rw_routines.f
> gfortran -O -w -ffixed-line-length-132 -c -o open_file.o open_file.f
> /home/app/gcc434/libexec/gcc/x86_64-unknown-linux-gnu/4.3.4/f951: error while loading shared libraries: libgmp.so.3: cannot open shared object file: No such file or directory
> /home/app/gcc434/libexec/gcc/x86_64-unknown-linux-gnu/4.3.4/f951: error while loading shared libraries: libgmp.so.3: cannot open shared object file: No such file or directory
> /home/app/gcc434/libexec/gcc/x86_64-unknown-linux-gnu/4.3.4/f951: error while loading shared libraries: libgmp.so.3: cannot open shared object file: No such file or directory
> make: /home/app/gcc434/libexec/gcc/x86_64-unknown-linux-gnu/4.3.4/f951: error while loading shared libraries: libgmp.so.3: cannot open shared object file: No such file or directory
> *** [open_file.o] Error 1
> make: *** Waiting for unfinished jobs....
> make: *** [ranmar.o] Error 1
> make: *** [gen_ximprove.o] Error 1
> make: *** [rw_routines.o] Error 1
>
> Please try to fix this compilations issue and retry.
> Help might be found at https://answers.launchpad.net/madgraph5.
> If you think that this is a bug, you can report this at https://bugs.launchpad.net/madgraph5
> ###################################
>
> It seems the library libgmp.so.3 cannot be found.
> But this library is indeed located at "/usr/lib/" and "/usr/lib64/".
>
> Can you help me link this lib ?
> Or is there any way I can fix this problem ?
>
> Thanks !
> Kechen
>
>
>
>
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Thank you for your reply, Olivier !

Do you mean I should run MG5 using the cluster mode instead ?
What I understand is:
I need to firstly to edit the card file Cards/me5_configuration.txt according to what you suggested.
Then, I input "/home/kechen/Python-2.6/python bin/generate_events" in a terminal (not qsub the shell code to the cluster) to generate events. MG5 will automatically submit jobs to cluster when it is generating events.
Is my understanding correct ?

I have one additional questions.
If I run in this way, what's the maximal number of events for a single run ?
Can I run more than 50,000 events for a single run ?

Thank you very much !
Kechen

Hi,

> Do you mean I should run MG5 using the cluster mode instead ?
yes

> I need to firstly to edit the card file Cards/me5_configuration.txt according to what you suggested.
> Then, I input "/home/kechen/Python-2.6/python bin/generate_events" in a terminal (not qsub the shell code to the cluster) to generate events. MG5 will automatically submit jobs to cluster when it is generating events.
> Is my understanding correct ?
yes

> I have one additional questions.
> If I run in this way, what's the maximal number of events for a single run ?
> Can I run more than 50,000 events for a single run ?

The limitation to the number of events is first due to pythia, where file larger than 4G can't handle correctly.
If you don't run pythia, you can, in principle, increase this number up to 1M.
But generating such many events make the unweighting/writting of events extremely slow.
So it might be better to use the multi_run options.

Cheers,

Olivier

On Nov 18, 2013, at 6:51 PM, Kechen Wang <email address hidden> wrote:

> Question #239418 on MadGraph5 changed:
> https://answers.launchpad.net/madgraph5/+question/239418
>
> Status: Answered => Open
>
> Kechen Wang is still having a problem:
> Thank you for your reply, Olivier !
>
> Do you mean I should run MG5 using the cluster mode instead ?
> What I understand is:
> I need to firstly to edit the card file Cards/me5_configuration.txt according to what you suggested.
> Then, I input "/home/kechen/Python-2.6/python bin/generate_events" in a terminal (not qsub the shell code to the cluster) to generate events. MG5 will automatically submit jobs to cluster when it is generating events.
> Is my understanding correct ?
>
> I have one additional questions.
> If I run in this way, what's the maximal number of events for a single run ?
> Can I run more than 50,000 events for a single run ?
>
> Thank you very much !
> Kechen
>
> --
> You received this question notification because you are a member of
> MadTeam, which is an answer contact for MadGraph5.

Olivier,

Now I am try generating events using "./madevent multi_run 10 testRun --laststep=pgs -f ".
It shows MG5 will automatically run 10 times for my processes.
For each run, it will also automatically submit jobs to the cluster.
This is good.

However, these runs have to be one by one.
I mean only after run_1 is finished, run_2 can run.
I found the pythia and PGS process are running using only one core.
This will take too long time for the next run.

Therefore, I have two questions:
(1) Is it possible that when run_1 is still running , run_2 can be started ?
      Something like running in parallel for multi-runs.
      Does it mean I can use the multi-core option (but I need to use the cluster. So I think I have to choose the cluster option) ?
(2) Can I run pythia and PGS using multi-core ?

Thanks !
Kechen

Thanks Olivier Mattelaer, that solved my question.

Thank you for your help, Olivier.
This helps me a lot.
Thanks !

Kechen