[Wien] segmentation fault error for beta-cristobalite

Jorissen Kevin Kevin.Jorissen at ua.ac.be
Thu Jul 15 16:49:08 CEST 2004


The memory required by lapw1 is approximately equal to
x  * nmat *  (nmat+num)
where nmat is the size of the hamiltonian (grep :RKM case.scf1) and num is the number of eigenvectors, and x is the size of one element of the matrices (eg. 8B for real, 16B for complex - I usually take x slightly larger to compensate for objects not considered in this simple rule etc.).
nmat is limited to the compilation parameter nmatmax, and num to nume.
 
So the formula above should help you reduce the limits nmatmax and nume and recompile so that you can be absolutely certain your computer is not requiring more memory than is available (limit it to 500 Mb, then you don't even need your swap space).
 
If this rids you of the segmentation fault, then you know that you don't have enough memory available to run the heavy case you were trying before (didn't look exceptionally heavy to me, though, but maybe you're requesting huge rkmax, ... ?).  Adding swap space like Torsten says, is one possibility, although for a machine with only 500 MB RAM, expanding the RAM is probably useful and still rather cheap (and of course a lot faster than relying strongly on swapping).
 
In any case, you should set nmatmax and nume so that the formula above does not exceed swap + RAM memory.  On a 2GB machine, I can run (for a real case with inversion symmetry) up to about nmatmax=12000 and nume=2000.  (for a complex case, divide nmatmax by sqrt(2)  You have only 1GB, so multiply again by sqrt(2)).
 
 
If  you can't get rid of the segmentation fault this way, recompile lapw1 with different compilation parameters in SRC_lapw1/Makefile.  eg. use -O0, try -static if you're not using that already, ...  Or try different libraries (mkl, default wien stuff, goto, ...).
 
 
If that doesn't help, ask again, and let's hope someone else has more useful ideas :-)
 
Good luck,
 
Kevin.
 
 

	-----Oorspronkelijk bericht----- 
	Van: Joel Abrahamson [mailto:jtabrahamson at yahoo.com] 
	Verzonden: wo 7/14/2004 11:39 
	Aan: wien at zeus.theochem.tuwien.ac.at 
	CC: 
	Onderwerp: [Wien] segmentation fault error for beta-cristobalite
	
	

	Dear Dr. Andersen,
	My total swap space is actually ~512 MB, with about
	256 MB each being listed as "used" and "free." The
	total memory of the computer is 512 MB, so I am not
	sure that swap space is the issue. I have also tried
	to decrease the size of the executables, but that has
	not helped either.
	Do you have any further suggestions?
	Sincerely,
	Joel Abrahamson
	^^^^^^
	Dear Mr. Abrahamson,
	
	if you have ensured that the shell limits are removed,
	try add some swap
	space - you might be running out of memory. Does it
	work with 1 (one)
	k-point?
	
	Best regards,
	Torsten Andersen.
	
	
	               
	__________________________________
	Do you Yahoo!?
	New and Improved Yahoo! Mail - Send 10MB messages!
	http://promotions.yahoo.com/new_mail
	
	_______________________________________________
	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: 6738 bytes
Desc: not available
Url : http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20040715/89c08e12/attachment.bin


More information about the Wien mailing list