[Wien] MPI parallelization

[Jorissen Kevin]

Hi Griselda,
if executing the command manually didn't work, the problem is either in your mpi/cluster configuration or in the command itself - maybe you need another one?  Have you succesfully used mpi for other applications?
Maybe your sys admin knows what's wrong?  
 
Kevin.
 

	


	The strange thing I saw is in the .processes file ... it is suppose that
	lapw1 is running with 18 processors but
	[griselda at clustersvr sd_v2]$ more .processes
	init:fisnode1
	1 : fisnode1 :  4 : 18 : 1
	
	or must be a similar file in each node changing just the name of the
	machine??
	
	
	The syntax of this file is : first you get one line for every machine the program is going to use.
	Then, for each job (just one in your case) there is a line giving : 
	number of the job : machine used : tail : number of processors used : number of the machine
	tail is related to the way the k-list has been split (depending on the weights of each processor); in your case I'd expect it to be the total number of k-points.
	I think just listing the name of the first machine is probably okay?  Your system will be told to start the lapw1_mpi from fisnode1 and use 18 processors.
	You can see that by looking at the command given by wien that you also tried manually.  the machine[$p] in that line is equal to 'fisnode1'.  And it should also be in your nohup.out by now.
	If this is NOT okay - if eg. all the names node1 to node9 have to be passed on somehow -, then you will have to change the $ttt line in lapw1para and replace it with the command that on your cluster will do what you want.  Ask sys admin for inspiration ...   Your nine nodes will be in $machine[1-9].
	 
	 
	MPI USERS CORRECT ME IF I'M WRONG ...
	 
	Kevin.
	 
	 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 5018 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20040429/e4f57b60/attachment.bin