Hi Nick,
According to the link you provided, we tried to compile siesta with FFLAGS =
-O0 -g -fbacktrace -fbounds-check, then it runs smoothly.It should be high
optimization caused segmention fault as you said. Thanks for your help!
Regards
Liang
-----原始邮件-----
发件人:"Nick Papior" <nickpap...@gmail.com>
发送时间:2017-11-17 18:03:59 (星期五)
收件人: "siesta-l@uam.es" <siesta-l@uam.es>
抄送:
主题: Re: [SIESTA-L] DM_out and segmentation error
Could you try and compile with debug options as specified in the previous link.
Perhaps this will shed some light on where and why this happens.
FYI I cannot reproduce this error on our cluster with your input, so I would,
currently, suspect a miscompilation.
Sadly I have too often seen intel being too aggressive in its optimizations and
thus resulting in seg-faults. Typically it can be leviated by reducing the
compiler optimization level for individual files (as you find for atom.F).
2017-11-16 4:00 GMT+01:00 王亮 <wangli...@cigit.ac.cn>:
Hello Nick,
I'm sorry for unclear describtion.Indeed, the fdf file I used is:
SystemName Bulk Silicon
SystemLabel Si
NumberOfSpecies 1
NumberOfAtoms 2
%block ChemicalSpeciesLabel
1 14 Si
%endblock ChemicalSpeciesLabel
LatticeConstant 5.43 Ang
%block LatticeVectors
0.00 0.50 0.50
0.50 0.00 0.50
0.50 0.50 0.00
%endblock LatticeVectors
AtomicCoordinatesFormat ScaledByLatticeVectors
%block AtomicCoordinatesAndAtomicSpecies
0.00 0.00 0.00 1
0.25 0.25 0.25 1
%endblock AtomicCoordinatesAndAtomicSpecies
%block kgrid_Monkhorst_Pack
4 0 0 0.5
0 4 0 0.5
0 0 4 0.5
%endblock kgrid_Monkhorst_Pack
Pseudopotontial is Si.psf lda ca nrl nc
In the header of output, we found:
Siesta Version: siesta-4.1--736
Architecture : unknown
Compiler flags: mpif90 -O2 -fPIC -ftree-vectorize -march=native
PP flags : -DFC_HAVE_ABORT -DMPI -DSIESTA__METIS -DSIESTA__MUMPS
Libraries :
-Wl,-rpath,/HOME/intel/composer_xe_2013_sp1.2.144/mkl/lib/intel64/
-L/lib/inntel64/ -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_core
-lmkl_sequential
-Wl,-rpath,/HOME/intel/composer_xe_2013_sp1.2.144/mkl/lib/intel64/
-L/mkl/lib/intel64/ -lmkl_blacs_intelmpi_lp64 -lmkl_blas95_lp64
-Wl,-rpath,/HOME/intel/composer_xe_2013_sp1.2.144/mkl/lib/intel64/
-L/lib/intel64/ -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_core
-lmkl_sequential /WORK/app/parmetis/4.0.3-N/lib/libmetis.a
/WORK/app/MUMPS/5.0.1/lib/libdmumps.a /WORK/app/MUMPS/5.0.1/lib/libcmumps.a
/WORK/app/MUMPS/5.0.1/lib/libmumps_common.a /WORK/app/MUMPS/5.0.1/lib/libpord.a
/WORK/app/MUMPS/5.0.1/lib/libsmumps.a /WORK/app/MUMPS/5.0.1/lib/libzmumps.a
-L/usr/local/mpi3-dynamic/lib/ -lmpicxx -lmpichf90 -lmpi -lmpifort -lstdc++
PARALLEL version
In the end, we found:
SCF cycle converged after 6 iterations
Using DM_out to compute the final energy and forces
No. of atoms with KB's overlaping orbs in proc 0. Max # of overlaps: 26
161
...forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
siesta 0000000000D55B59 Unknown Unknown Unknown
siesta 0000000000D5442E Unknown Unknown Unknown
siesta 0000000000CF8B52 Unknown Unknown Unknown
siesta 0000000000CA8CC3 Unknown Unknown Unknown
siesta 0000000000CAEFCB Unknown Unknown Unknown
libpthread.so.0 000000392220F710 Unknown Unknown Unknown
siesta 00000000007AF9E9 Unknown Unknown Unknown
siesta 000000000081DCCC Unknown Unknown Unknown
siesta 000000000063AD03 Unknown Unknown Unknown
siesta 0000000000680E8D Unknown Unknown Unknown
siesta 0000000000BB822F Unknown Unknown Unknown
siesta 000000000042AF16 Unknown Unknown Unknown
libc.so.6 0000003921A1ED1D Unknown Unknown Unknown
siesta 000000000042AE09 Unknown Unknown Unknown
This fault appeared whether ran on 1 cpu or 24 cpus.
Our platform is MILKYWAY-2 - TH-IVB-FEP cluster.
operatings ystem:
LSB Version:
:base-4.0-amd64:base-4.0-noarch:core-4.0-amd64:core-4.0-noarch:graphics-4.0-amd64:graphics-4.0-noarch:printing-4.0-amd64:printing-4.0-noarch
Distributor ID: RedHatEnterpriseServer
Description: Red Hat Enterprise Linux Server release 6.5 (Santiago)
Release: 6.5
Codename: Santiago
compiler：
gcc： Thread model: posix
gcc version 4.4.7 20120313 (Red Hat 4.4.7-4) (GCC)
mpif90：
mpifort for MPICH version 3.1.3
ifort version 14.0.2
You can also find arch.make file in attachment.
Thanks!
Best regards!
-----原始邮件-----
发件人:"Nick Papior" <nickpap...@gmail.com>
发送时间:2017-11-15 15:56:59 (星期三)
收件人: "siesta-l@uam.es" <siesta-l@uam.es>
抄送:
主题: Re: [SIESTA-L] DM_out and segmentation error
Could you please follow the instructions on this page:
https://answers.launchpad.net/siesta/+faq/2779
This would help in tracking down the problem.
2017-11-14 13:17 GMT+01:00 王亮 <wangli...@cigit.ac.cn>:
Hello,
I tried examples from siesta official page （Files-Latcon-SrTiO3，
File-Si-structure...）,after scf, I always got segmentation error:
Using DM_out to compute the final energy and forces
No. of atoms with KB's overlaping orbs in proc 0. Max # of overlaps: 26
161
...forrtl: severe (174): SIGSEGV, segmentation fault occurred
Does anyone know how to fix it?
Ps: siesta-4.1b3 and 24 cpus or 1 cpu
Regards
--
Liang WANG
Center for Nanofabrication and System Integration
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of
Science
266 Fangzheng Ave, ShuiTu technology development zone, Beibei District,
Chongqing, P. R. China
Tel: +86-23-65935570
--
Kind regards Nick
--
Liang WANG
Center for Nanofabrication and System Integration
Chongqing Institute of Green and Intelligent Technology, Chinese Academy of
Science
266 Fangzheng Ave, ShuiTu technology development zone, Beibei District,
Chongqing, P. R. China
Tel: +86-23-65935570
--
Kind regards Nick
