forum@abinit.org
Subject: The ABINIT Users Mailing List ( CLOSED )
List archive
- From: Matthieu Verstraete <mjv500@york.ac.uk>
- To: forum@abinit.org
- Subject: mac OSX mpi error
- Date: Tue, 9 May 2006 19:58:07 +0100 (BST)
Hello all,
with the latest 5.1.3 abinip version on our cluster of mac G5 machines running mac OSX, I get an error during the first cycle. I suspect it's when the mother node passes the wavefunctions to the others. The error can show up as
1) an error message claiming the code wasn't able to allocate (roughly) 600 Mb (it should be able to)
2) an MPI error claiming it tried to pass a 600 Mb message, larger than the maximum mpi message size.
3) an MPI p4 error:
p1_21744: p4_error: net_recv read: probable EOF on socket: 1
I think the crash is in vtorho, and that is also where the calls to mpi_recv are for the wfk etc...
With version 4.5.2, no error occurs. Someone has been playing around with the mpi calls! Naughty. And can anyone fix them?
My input is concatenated below (for Masayoshi :) but it's a huge job.
You need the GGA TM pseudopotentials for Ag and Pd.
Has anyone else had this? It could be platform specific, or a memory/size thing.
Matthieu
PS: thanks to the Paris guys who coded iscf 7. It's a revolution (we need a few more revolutions like that).
--
================================================================
Dr. Matthieu Verstraete mailto:mjv500@york.ac.uk
Dept. of Physics, University of York, tel: +44 1904 43 22 08
Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14
STARTINPUT STARTINPUT STARTINPUT STARTINPUT STARTINPUT STARTINPUT
ndtset 2
mkmem 0
getwfk2 1
prtstm2 1
stmbias2 -0.5 eV # Might have been given in Ha, but the eV units are more common
nstep2 1
tolwfr2 1.0d-6
occopt2 7
tsmear2 0.01
#Definition of the unit cell 1 couche = 4.32 bohr
acell 21.528500 5.382125 51.84
xcart
2.6910637E+00 1.5536857E+00 -4.3467002820 # Ag
0.0000000000 0.0000000000 1.2709597486E-01 # Pd
0.0000000000 3.1073715E+00 4.5392706055E+00 # Pd
2.6910637E+00 1.5536857E+00 8.9514452361E+00 # Pd
0.0000000000 0.0000000000 1.3425241493E+01 # Ag
8.0731887 1.5536857E+00 -4.3467002820 # Ag
5.3821250000 0.0000000000 1.2709597486E-01 # Pd
5.3821250000 3.1073715E+00 4.5392706055E+00 # Pd
8.0731887 1.5536857E+00 8.9514452361E+00 # Pd
5.3821250000 0.0000000000 1.3425241493E+01 # Ag
13.4553137 1.5536857E+00 -4.3467002820 # Pd
10.7642500000 0.0000000000 1.2709597486E-01 # Pd
10.7642500000 3.1073715E+00 4.5392706055E+00 # Pd
13.4553137 1.5536857E+00 8.9514452361E+00 # Pd
10.7642500000 0.0000000000 1.3425241493E+01 # Pd
18.8374387 1.5536857E+00 -4.3467002820 # Pd
16.1463750000 0.0000000000 1.2709597486E-01 # Pd
16.1463750000 3.1073715E+00 4.5392706055E+00 # Pd
18.8374387 1.5536857E+00 8.9514452361E+00 # Pd
16.1463750000 0.0000000000 1.3425241493E+01 # Pd
#Optimization of the lattice parameters
#Definition of the planewave basis set
ecut 50.0
kptopt 1
nband 180
nshiftk 1
shiftk 0.0 0.0 0.5
kptrlatt 4 0 0 0 8 0 0 0 1
#Definition of the SCF procedure
nstep 200 # Maximal number of SCF cycles
iscf 7
iprcel 45
cpuh 90
tolvrs 1.0d-5
#Definition of occupation numbers
occopt 4
tsmear 0.03
ixc 11
angdeg 90 90 120
diemix 0.3
#Definition of the atom types
ntypat 2
znucl 46 47
#Definition of the atoms
natom 20
typat
2 1 1 1 2
2 1 1 1 2
1 1 1 1 1
1 1 1 1 1
- mac OSX mpi error, Matthieu Verstraete, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, james raynolds, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, Matthieu Verstraete, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, james raynolds, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, Matthieu Verstraete, 05/10/2006
- Re: [abinit-forum] mac OSX mpi error, james raynolds, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, Matthieu Verstraete, 05/09/2006
- Re: [abinit-forum] mac OSX mpi error, james raynolds, 05/09/2006
Archive powered by MHonArc 2.6.16.