[Wien] BACKSPACE error in lapwso calculation

[Md. Fhokrul Islam]

Hi Gavin,


I have compiled Wien2k with -O0 option but it still failed with BACKSPACE error at LAPWSO.

I have tried to run the same calculation in a different cluster where I have compiled Wien2k14.2

with intel/2016 and looks like this error message doesn't appear. Although the job crashed with

virtual memory error but at least it produced some output files from LAPWSO before it crashed.

SO I guess the problem is with intel/2017 compiler.


Thanks again for your reply.


Regards,

Fhokrul



________________________________
From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at> on behalf of Gavin Abo <gsabo at crimson.ua.edu>
Sent: Wednesday, March 22, 2017 5:20 AM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] BACKSPACE error in lapwso calculation

I cannot say for sure, but I would suspect that it is caused by an Intel Fortran bug.

Previously, you reported an "end-of-file during read" error [1] with intel/2017 [2].  And now, you report "BACKSPACE error", which is another file reading error.  Maybe the errors observed in 2016 Update 3 [3] haven't entirely been resolved in intel/2017.

Did this error occur when WIEN2k was compiled with -O0?  If not, I suggest trying it to see if the error goes away or not.

Version 15 had a BACKSPACE bug in the run-time library (RTL) [4].  So if there is a bug, it might come from this library.  Have you tried the calculation when WIEN2k was compiled with the dynamic libraries, then again with the static libraries using the Intel link advisor [5]?  If not, you might give that a try.  I think I encountered a Fortran bug once in one of the buggy 12 versions were an error occurred with the dynamic libraries, but not with the static ones.  Though, maybe it won't help in this case, since it looks like the static and dynamic linking uses the same library file (libiomp5.so) [6].

How large is the 1Mn.vectordn_1 file in size?  Do these errors only occur with calculations with large files sizes and smaller file size calculations using the same sequence of WIEN2k program commands work fine?  Perhaps it could be a big data issue if that is the case.  I don't remember where I saw it, but I thought I read somewhere that when there are clusters and big data are involved, they can be more complex to handle.

I don't know if there is a boundary limit on the amount of data BACKSPACE can handle [7].  I'm not able to see from the Intel documentation whether say a small BLOCKSIZE [8,9] could slow down the read, perhaps lead to a time out, and cause such an error.  Also, it is not clear whether or not it could be caused by a memory leak or memory limit from BACKSPACE using n - 1 successive READs [10, 11]. It might help to ask the experts about the error on the Intel forum [12].

[1] http://zeus.theochem.tuwien.ac.at/pipermail/wien/2017-March/026101.html
[2] http://zeus.theochem.tuwien.ac.at/pipermail/wien/2017-March/026106.html
[3] https://software.intel.com/en-us/articles/read-failure-unformatted-file-io-psxe-16-update-3

[4] https://software.intel.com/en-us/forums/intel-visual-fortran-compiler-for-windows/topic/549033

[5] https://software.intel.com/en-us/articles/intel-mkl-link-line-advisor

[6] https://software.intel.com/en-us/node/528514

[7] https://software.intel.com/en-us/forums/intel-visual-fortran-compiler-for-windows/topic/274239

[8] http://kiwi.atmos.colostate.edu/rr/tidbits/intel/macintel/doc_files/source/extfile/optaps_for/fortran/optaps_prg_io_f.htm

[9] https://software.intel.com/en-us/node/678835

[10] https://software.intel.com/en-us/forums/intel-visual-fortran-compiler-for-windows/topic/593848

[11] https://software.intel.com/en-us/node/692589?language=en
[12] https://software.intel.com/en-us/forums/intel-fortran-compiler-for-linux-and-mac-os-x


On 3/21/2017 8:46 AM, Md. Fhokrul Islam wrote:
Hi Prof Blaha and Wien2k users,

I am running a surface supercell calculation with spin-orbit interaction using non-scf
method.  While LAPW1 worked fine and produced the energy and scf1 files correctly,
LAPWSO failed with BACKSPACE error. Please see the error message below. I have checked
the get_nloat.f file and while there is a BACKSPACE statement in line 22 but it is not
obvious to me why there is a problem to execute this line. Somewhere it seems to have
generated an 'forrtl: Invalid argument' after LAPW1 but not sure why. I am using Wien2k_16
which I have compiled with intel/2017.

 LAPW1 END
 LAPW1 END
 LAPW1 END
 LAPW1 END
 LAPW1 END
 LAPW1 END
 LAPW1 END
 LAPW1 END
forrtl: Invalid argument
forrtl: severe (23): BACKSPACE error, unit 9, file /scratch/1Mn.vectordn_1
Image              PC                Routine            Line        Source
lapwso             0000000001D5AFBC  Unknown               Unknown  Unknown
lapwso             0000000001D57771  Unknown               Unknown  Unknown
lapwso             00000000004305D6  get_nloat_                 22  get_nloat.f
lapwso             0000000000419F2F  MAIN__                    144  lapwso.F
lapwso             000000000040442E  Unknown               Unknown  Unknown
libc-2.23.so       00002AF2B4B9A830  __libc_start_main     Unknown  Unknown
lapwso             0000000000404329  Unknown               Unknown  Unknown

Any suggestion will be appreciated.

Thanks,
Fhokrul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20170324/9207f261/attachment-0001.html>