[Smeagol-discuss] segmentation fault
Barraza-lopez, Salvador
sbl3 at mail.gatech.edu
Tue Apr 7 15:27:06 IST 2009
Serhan,
Do you have a way to run this calculation in parallel? The options you might have now is to reduce the number of atoms in your calculation. I would be concerned about the precision with such small cutoffs and small basis set.
Another thing, the memory is allocated dynamically in smeagol. If you plot and look very finely at the output from the free command I mentioned a number of weeks back, you might see that the memory in the code peaks sharply at some points during execution. My point here is that if you see the memory well contained before the crash, you still have to trust your seg fault message: The memory peaks until a point where the program crashes.
Best regards,
-Salvador.
----- Original Message -----
From: "Ivan Rungger" <runggeri at tcd.ie>
To: "Serhan Yamacli" <syamacli at gmail.com>
Cc: smeagol-discuss at lists.tchpc.tcd.ie
Sent: Tuesday, April 7, 2009 10:13:50 AM GMT -05:00 US/Canada Eastern
Subject: Re: [Smeagol-discuss] segmentation fault
Hello Serhan,
it might still be a problem of memory. Using different compilers will
change a bit the way the memory space is allocated, but at the end of
the day if one runs out of memory it is likely that also the other one
runs out of memory. So probably the newly compiled code allocates memory
in a somewhat more efficient way, but at the end it also runs out of
memory. There might also be library problems, but in this case I think
it is still the memory.
Cheers,
Ivan
Serhan Yamacli wrote:
> Hi Ivan and other friends,
>
> I've tried the simulation using Single Zeta basis set, the program
> advanced a bit, but it again gave segmentation error like:
>
> -------------------------------------------
> InitMesh: Mesh cutoff (required, used) = 100.000 103.137 Ry
>
> * Maximum dynamic memory allocated = 62 MB
>
> (NOTE: PREVIOUSLY, THE PROGRAM GAVE SEGMENTATION FAULT HERE, BUT IT
> CONTINUED NOW)
>
> gensvd: Leads decimation
> gensvd: Dim of H1 and S1 : 324
> gensvd: Rank of H1: 38
> gensvd: Rank of (H1,S1): 117
> gensvd: Decimated states: 169
> gensvd: Decimation from the left
> gensvd: Leads decimation
> gensvd: Dim of H1 and S1 : 324
> gensvd: Rank of H1: 38
> gensvd: Rank of (H1,S1): 117
> gensvd: Decimated states: 169
> gensvd: Decimation from the left
> forrtl: severe (174): SIGSEGV, segmentation fault occurred
> Image PC Routine Line
> Source
> libblas.so.3 00002B174D0DA380 Unknown Unknown
> Unknown
> liblapack.so.3 00002B174CC2B35C Unknown Unknown
> Unknown
> liblapack.so.3 00002B174CC78B4C Unknown Unknown
> Unknown
> smeagol 000000000061548B Unknown Unknown
> Unknown
> smeagol 00000000005F6028 Unknown Unknown
> Unknown
> smeagol 000000000065D65F Unknown Unknown
> Unknown
> smeagol 0000000000663F8D Unknown Unknown
> Unknown
> smeagol 0000000000599F0B Unknown Unknown
> Unknown
> smeagol 000000000040546C Unknown Unknown
> Unknown
> libc.so.6 00002B174D7A0586 Unknown Unknown
> Unknown
> smeagol 0000000000405369 Unknown Unknown
> Unknown
> --------------------------------------------------------------------------
>
> Is this a problem about libraries?
>
> Cheers,
> Serhan
>
> 03 Nisan 2009 Cuma 13:50 tarihinde Ivan Rungger <runggeri at tcd.ie
> <mailto:runggeri at tcd.ie>> yazdı:
>
> Hello Serhan,
>
> smeagol can in principle be compiled with gfortran, but I don't know.
> The ifort compiler is surely the most used one, so with that one the
> code should work well. Anyway also to compile with gfortran it
> would be
> good to have compatible libraries preinstalled. If you compile the
> code
> in serial mode, than you can use the libraries provided in the
> smeagolpack directory, to do this just edit the Makefile accordingly
> (some lines have to be commented out, and some others commented I
> think).
>
> Can you please send the debug error message for the single zeta
> run, and
> also all the standard output file? Because the last debug error
> you sent
> clearly showed that the segmentation fault occurred when a large piece
> of memory was allocated. So if you resend the error message for the SZ
> run, we can see if there is the same problem. A note: just make sure
> that you set all the limits related to memory (especially the stack
> size) to unlimited in your shell when you run the code.
>
> Cheers,
>
> Ivan
>
> Serhan Yamacli wrote:
> > Hi Ivan,
> >
> > I've used single zeta basis set but it gave the same error. Now, I'm
> > trying to install Smeagol again with gfortran compiler. Which
> > arch.make can I use with gfortran without any libraries installed to
> > the system?
> >
> > Cheers,
> > Serhan
> >
> >
> > 2009/4/3 Ivan Rungger <runggeri at tcd.ie <mailto:runggeri at tcd.ie>
> <mailto:runggeri at tcd.ie <mailto:runggeri at tcd.ie>>>
> > - Alıntı metni gizle -
> >
> > Hello Serhan,
> >
> > Do you have the debugging error output for this run? I think it
> might
> > still be a memory issue. Did/Can you try to get the single zeta
> > calculation to run, because that should use a lot less memory?
> >
> > Cheers,
> >
> > Ivan
>
>
> --
> =================================================
> Ivan Rungger,
>
> School of Physics and CRANN,
> Trinity College Dublin,
> Dublin 2, IRELAND
> Phone: +353-1-8968454
> Email: runggeri at tcd.ie <mailto:runggeri at tcd.ie>
> =================================================
>
>
>
--
=================================================
Ivan Rungger,
School of Physics and CRANN,
Trinity College Dublin,
Dublin 2, IRELAND
Phone: +353-1-8968454
Email: runggeri at tcd.ie
=================================================
_______________________________________________
Smeagol-discuss mailing list
Smeagol-discuss at lists.tchpc.tcd.ie
http://lists.tchpc.tcd.ie/listinfo/smeagol-discuss
--
Salvador Barraza-Lopez
Postdoctoral Fellow
School of Physics
The Georgia Institute of Technology
Office N205
837 State Street Atlanta, Georgia 30332-0430 U.S.A
Tel: (404) 894-0892 Fax: (404) 894-9958
More information about the Smeagol-discuss
mailing list