[Smeagol-discuss] segmentation fault
Barraza-lopez, Salvador
sbl3 at mail.gatech.edu
Sun Mar 22 12:37:01 GMT 2009
Hi Serhan,
If you want to have a closer look at memory try:
/usr/bin/free -m -s 0.5 >> memfile
This will look for (m)memory every 0.5 seconds and dump output to file memfile.
Once the program crashes, have a look at your last line on memfile (top can be too coarse). This will be the best way to actually know how much memory you're using.
-Salvador.
----- Original Message -----
From: "Serhan Yamacli" <syamacli at gmail.com>
To: "Salvador Barraza-lopez" <sbl3 at mail.gatech.edu>, smeagol-discuss at lists.tchpc.tcd.ie
Sent: Sunday, March 22, 2009 4:44:43 AM GMT -05:00 US/Canada Eastern
Subject: Re: [Smeagol-discuss] segmentation fault
Hi,
I compiled with these options. And then googled for the error and found that it's a memory error as you said. I increased the stack size by the command ulimit -s unlimited. The program now does not give segmentation error but instead gives the "insufficient virtual memory" error. I've also attached the screenshot of the system resources at their maximum when running Smeagol. Now the problem is that Smeagol uses 165MB of RAM at maximum (just before it gives error) but I have 2GB of RAM with 4GB of swap. It can not gather the remaining RAM which is more then 1GB and gives the following error:
----------------------------------------------------------------------------------------------------------------
* Maximum dynamic memory allocated = 174 MB
gensvd: Leads decimation
gensvd: Dim of H1 and S1 : 468
gensvd: Rank of H1: 49
gensvd: Rank of (H1,S1): 154
gensvd: Decimated states: 265
gensvd: Decimation from the left
gensvd: Leads decimation
gensvd: Dim of H1 and S1 : 468
gensvd: Rank of H1: 49
gensvd: Rank of (H1,S1): 154
gensvd: Decimated states: 265
gensvd: Decimation from the left
forrtl: severe (41): insufficient virtual memory
Image PC Routine Line Source
smeagol 0831F29D Unknown Unknown Unknown
smeagol 0831E815 Unknown Unknown Unknown
smeagol 082DB838 Unknown Unknown Unknown
smeagol 0829BB43 Unknown Unknown Unknown
smeagol 082C4380 Unknown Unknown Unknown
smeagol 081CEDAC negf_ 570 negf.F
smeagol 082679E3 emtg_ 280 emtg.F
smeagol 0826DE62 emtrans_ 208 emtrans.F
smeagol 08194A98 MAIN__ 1698 siesta.F
smeagol 0804B4D1 Unknown Unknown Unknown
libc.so.6 40B5A685 Unknown Unknown Unknown
smeagol 0804B3E1 Unknown Unknown Unknown
---------------------------------------------------------------------------------------------------------------------
Thanks again.
Cheers
Serhan
2009/3/21 Barraza-lopez, Salvador < sbl3 at mail.gatech.edu >
When you compile it, use -O0 -g -traceback. I guess you are using gcc?
My gut feeling is you don't have sufficient memory to run it... The important number here is
initatomlists: Number of atoms, orbitals, and projectors: 60 540 960
the number of orbitals. You say it decreased when you changed your basis size. The 540 is 1s, 3p and 5d per 60 atoms (9*60=540). It went down from 13*60= 780 that you had before... In any case, try running it in debug mode and use top at the same time.
as an additional test, try SZ for your basis, that'll give you 4 orbitals per atom, matrices of size 240x240. If it crashes, try compiling in debug mode.
Also, investigate in google what segmentation fault means. I do believe is a memory related issue.
Best,
-Salva.
----- Original Message -----
From: "Serhan Yamacli" < syamacli at gmail.com >
To: "Salvador Barraza-lopez" < sbl3 at mail.gatech.edu >
Sent: Saturday, March 21, 2009 1:48:02 PM GMT -05:00 US/Canada Eastern
Subject: Re: [Smeagol-discuss] segmentation fault
I'm using Smeagol at a single machine in serial mode. The CPU is 2.26GHz and memory available is 2GB. How can I run it in debug mode?
Cheers
Serhan
2009/3/21 Barraza-lopez, Salvador < sbl3 at mail.gatech.edu >
can you run it in debug mode? can you share the output? how much memory do you have available per processor?
Best,
-Salvador.
----- Original Message -----
From: "Serhan Yamacli" < syamacli at gmail.com >
To: "Salvador Barraza-lopez" < sbl3 at mail.gatech.edu >, smeagol-discuss at lists.tchpc.tcd.ie
Sent: Saturday, March 21, 2009 11:59:38 AM GMT -05:00 US/Canada Eastern
Subject: Re: [Smeagol-discuss] segmentation fault
Hi Salvador,
I looked at the system resources as you said. Smeagol executable takes 26% of memory at maximum, then it gives the segmentation fault error.
Best regards,
Serhan
2009/3/21 Barraza-lopez, Salvador < sbl3 at mail.gatech.edu >
Hi Serhan,
Can you check the dynamic memory in your run? I believe the problem is related to this but I may be wrong.
you can log in to the running node, type top and get an idea of the total memory as the program goes.
Best regards,
-Salvador.
----- Original Message -----
From: "Serhan Yamacli" < syamacli at gmail.com >
To: "Salvador Barraza-lopez" < sbl3 at mail.gatech.edu >
Sent: Saturday, March 21, 2009 11:19:03 AM GMT -05:00 US/Canada Eastern
Subject: Re: [Smeagol-discuss] segmentation fault
Hi Salvador,
Thanks for the answer but, Smeagol again gives the error when changed basis to SZP:
-----------------------------------------------------------------------------------------
initatomlists: Number of atoms, orbitals, and projectors: 60 540 960
siesta: System type = chain
* Maximum dynamic memory allocated = 1 MB
siesta: ===============================
SMEAGOL Bias step = 0, V = 0.000 Ry
Begin CG move = 0
===============================
InitMesh: MESH = 180 x 180 x 108 = 3499200
InitMesh: Mesh cutoff (required, used) = 200.000 212.441 Ry
* Maximum dynamic memory allocated = 279 MB
Segmentation fault
-------------------------------------------------------------------------------------
I again attached the modidied input fdf.
Cheers,
Serhan
2009/3/21 Barraza-lopez, Salvador < sbl3 at mail.gatech.edu >
Hi Serhan, how about reducing the basis size to SZP?
Best regards,
-Salvador.
----- Original Message -----
From: "Serhan Yamacli" < syamacli at gmail.com >
To: smeagol-discuss at lists.tchpc.tcd.ie
Sent: Saturday, March 21, 2009 10:34:06 AM GMT -05:00 US/Canada Eastern
Subject: [Smeagol-discuss] segmentation fault
Hi,
I'm trying to simulate a (3,3) intrinsic carbon nanotube. I performed the leads calculation. But when I try to simulate the transport, Smeagol says:
--------------------------------------------------------------------------------------------------------------
initatomlists: Number of atoms, orbitals, and projectors: 60 780 960
siesta: System type = chain
* Maximum dynamic memory allocated = 1 MB
siesta: ===============================
SMEAGOL Bias step = 0, V = 0.000 Ry
Begin CG move = 0
===============================
InitMesh: MESH = 240 x 240 x 150 = 8640000
InitMesh: Mesh cutoff (required, used) = 400.000 409.800 Ry
* Maximum dynamic memory allocated = 727 MB
Segmentation fault
-------------------------------------------------------------------------------------------------------
My fdf file is atteched. How can I solve this problem? I'm using Mandriva 2008 with ifort.
Cheers
Serhan
_______________________________________________ Smeagol-discuss mailing list Smeagol-discuss at lists.tchpc.tcd.ie http://lists.tchpc.tcd.ie/listinfo/smeagol-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.tchpc.tcd.ie/pipermail/smeagol-discuss/attachments/20090322/25c45804/attachment-0001.html
More information about the Smeagol-discuss
mailing list