TELEMAC-MASCARET Forum: running v8p0r0 in parallel (2/2)

running v8p0r0 in parallel 5 years 10 months ago #32677

olslewfoot
OFFLINE
Senior Boarder
Posts: 132
Thank you received: 3

Hi Yugi
I can now get partly through the scaler api build but it fails at the end of first part of f2py at api 127.

For the parallel runs other than add your suggested keys for gfortran and f2py3 the compilation doesn't start - Could not find cmdf file for config - use the --rescan option

I have tried --rescan - but no success. Error text attached.
Can you suggest anything?
John

Attachments:

compile_er...pi_2.txt (6KB)

The administrator has disabled public write access.

running v8p0r0 in parallel 5 years 10 months ago #32679

yugi OFFLINE openTELEMAC Guru Posts: 851 Thank you received: 244	for the f2py the ekyword needs to contains the path to your f2py for python3. For me it was f2py3 but sometimes it is f2py3.4, f2py, f2py3.7 ... have a look at which one you have. The other error with the cmdf is stranger. Maybe replace in your .cfg: modules; update -mascaret by modules: system -mascaret
	There are 10 types of people in the world: those who understand binary, and those who don't. The administrator has disabled public write access.

running v8p0r0 in parallel 5 years 10 months ago #32680

olslewfoot OFFLINE Senior Boarder Posts: 132 Thank you received: 3	Hi Yugi Some success - f2py was the correct key thanks - so the two parallel configs will now complete successfully but the scaler build has an error during the second part of f2py for api 1. Ive attached the error script. Would be grateful for any further suggestions? Attachments: compile_er...pi_3.txt (12KB)
	The administrator has disabled public write access.

running v8p0r0 in parallel 5 years 10 months ago #32681

yugi OFFLINE openTELEMAC Guru Posts: 851 Thank you received: 244	Looks like the api (fortran part) was not compiled. If you have a look in <root>/builds/<USETELCFG>/wrap_api/lib/ i am guessing that libapi.so is not in it. It should also not be in <root>/builds/<USETELCFG>/lib/ named api.so Could be because of your [shared] configuration. I have not tested that fonctionality that sebastien is using for nag_ad and other things.
	There are 10 types of people in the world: those who understand binary, and those who don't. The administrator has disabled public write access. The following user(s) said Thank You: olslewfoot

running v8p0r0 in parallel 5 years 10 months ago #32682

olslewfoot
OFFLINE
Senior Boarder
Posts: 132
Thank you received: 3

Thanks Yugi

The scalar config runs successfully although the api has not compiled although it will not do so unless I add the mpi_cmdexec key? then it will run successfully.

I have an error somewhere in the parallel runs as hpc_stdin gets into a loop - but this must be in the hpc_stdin commands. I have used the same commands as Sebstien suggested (and have used on my cluster before) - but something is wrong. I will find it.

Thank you for your considerable time and help.
John

The administrator has disabled public write access.

running v8p0r0 in parallel 5 years 9 months ago #32688

yugi
OFFLINE
openTELEMAC Guru
Posts: 851
Thank you received: 244

For the submit it again an error in python3.
I found out about it last week.
And have made the corrections on the trunk.
The issue is that the option --mpi is ignored.
You can have a look in the trunk at the correction it was in commit 12706
The code below was the correction

@@ -549,9 +549,9 @@ class Study(object):
         #     - runcmd if options.hpc
         #     - cas_files[name]['run'] and cas_files[name]['sortie'] otherwise
         # update mpi command if necessary
-        if self.cfg['MPI'] != {}:
+        if self.cfg['MPI'] != {} or options.mpi:
             self.run_cmd = self.mpi_cmd.replace('<exename>', self.exe_name)
-        if self.cfg['HPC'] == {}:
+        if self.cfg['HPC'] == {} or options.mpi:
             self.run_local()
 
         # <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

There are 10 types of people in the world: those who understand binary, and those who don't.

The administrator has disabled public write access.

running v8p0r0 in parallel 5 years 9 months ago #32691

olslewfoot OFFLINE Senior Boarder Posts: 132 Thank you received: 3	Thanks Yugi - the loop problem solved with the amendment in study.py Thanks you for your patience and help with this thread. JOhn
	The administrator has disabled public write access.