[Wien] .machines file

[Jorissen Kevin]

Hi, 
I think that in your case, I'd distribute the k-points using k-points parall. first, i.e., make a machines-file containing four lines like 
1: {machines to use for this k-point}
 
Since 8 processors is already quite something, and 48 atoms is not so horrible, maybe just one node per k-point?  So, sth like
1: node1 : 8
...
1:node4 : 8
 
This will use 8 processors on one node for each of the four k-points.
 
Of course, if the network is fast and you don't need the other nodes for other things, you can add more.
 
And then add some more lines to the machines-file, like granularity:1, and lapw0:{some machines}, and maybe extrafine ...
 
 
You can ignore the hup-error ; if it bothers you, just find the line in the beginning of run_lapw or runsp_lapw or other scripts where the hup-command is given, and delete that line (it's not necessary).
The 'lapw1cpara'-command not found IS a problem : is your $WIENROOT-variable configured correctly?  On all the machines?
Go look in the /paci/ucsd/u11341/WIEN/inst/ directory and see whether you can find lapw1cpara there, and check whether it's available with the same name on all your machines.
 
Hope this helps you ...
 
Kevin.
 

	 
	 
	-----Original Message----- 
	From: Griselda Garcia [mailto:griselda at bethe.fis.puc.cl] 
	Sent: Thu 8/28/2003 8:54 PM 
	To: wien at zeus.theochem.tuwien.ac.at 
	Cc: 
	Subject: [Wien] .machines file
	
	

	Dear WIEN users,
	
	Can anybody explain me the differences in .machines file when the
	k-point or fine grained parallelization is used?
	I read teh WIEN manual but I di not understand how to write the
	.machines file for my problem.
	I want to use an IBM terascale machine (arranged as 144 nodes and each
	node is equipped with 8 Power3 processors)
	to calculate a system of 48 nonequivalent atoms and 4 k-points in the BZ.
	I have to use a script to send my job to a LoadLeveler system, which
	distribute the jobs in the computer, so I do not know which processor
	will be assign to me.
	
	I did an attempt but I failed. Below I attach the files used to send my
	job to the LoadLeveler system, the error file at the end of the job to
	get some hints from you.
	
	The script to send my job:
	#!/usr/bin/csh
	#@environment = COPY_ALL; MP_EUILIB=us;
	XLSMPOPTS=parthds=1:spins=0:yields=0:schedule=static; AIXTHREAD_SCOPE=S;
	AIXTHREAD_MNRATIO=8:8;MP_SHARED_MEMORY=yes;MP_PULSE=0;
	#@Job_name=si111   
	#@job_type=parallel
	#@network.MPI = csss,shared,US
	#@output=$(job_name).$(jobid).out
	#@error=$(job_name).$(jobid).err
	#@node=6
	#@tasks_per_node=8
	#@node_usage = shared
	#@notify_user=ggarcia at fis.puc.cl
	#@notification=error
	#@notification=complete
	#@wall_clock_limit=10:00:00
	#@class=normal
	#@queue
	
	cat << EOF >.machines
	granularity:1
	1:bluehorizon.npaci.edu:4
	lapw0:bluehorizon.npaci.edu:1
	EOF
	runsp_lapw -p
	
	
	The error files sent after my job ran:
	
	stty: tcgetattr: A specified file does not support the ioctl system call.
	hup: Command not found.
	STOP  LAPW0 END
	/paci/ucsd/u11341/WIEN/inst/lapw1cpara: Command not found.
	
	
	Thank you so much!!!
	
	Regards, Griselda
	
	
	_______________________________________________
	Wien mailing list
	Wien at zeus.theochem.tuwien.ac.at
	http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
	

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 6914 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20030829/210b06bc/attachment.bin