File too large

[tolsen]

Hi

I get the error message when running a BSE calculation on MoS2 with 10x30x30 kpoints (846 irreducible).


.
.
.
<---> [FFT-BSK] Mesh size: 12 12 45
<---> [WF-BSK loader] Wfs (re)loading | | [000%] --(E) --(X)
<---> [WF-BSK loader] Wfs (re)loading |####################| [100%] --(E) --(X)
<---> [M 1.192 Gb] Alloc BS_W (0.167)
<---> [05.01] Main loop


<---> BSK | | [000%] --(E) --(X)
P01: [ERROR] STOP signal received while in :[05.01] Main loop
P01: [ERROR][NetCDF] File too large



It seems to work fine when I perform the calculation on the system using less kpoints. I use the -S option. What file becomes too large? And is there anyway around this problem?

BR
Thomas Olsen
Post Doc
Technical University of Denmark

[andrea marini]

This is strange. The -S, especially in the case of many k-points, should fragment the BS kernel in small pieces. Can you check the the code is actually creating the fragments in the SAVE folder ? Otherwise please post the GS input file so that we can try to reproduce the error.

Andrea

[myrta gruning]

Hallo Thomas

Besides checking that the BS matrix is fragmented, check the SAVE dir for files larger than 2Gb. Attach also the input that may help us to see what can cause the error.
Best,
m

[tolsen]

Hi

There is lots of BS fragments in the SAVE directory. The calculation fails while at the ndb.BS_Q1_fragments_112_744 fragment. I guess it should go up to ndb.BS_Q1_fragments_001_846. The RESTART/db file writes:

Section Completed 277030 . To reach 358282

I tried to attach the yambo.in file but it was told that the extension is not allowed.

BR
Thomas

[myrta gruning]

There is lots of BS fragments in the SAVE directory. The calculation fails while at the ndb.BS_Q1_fragments_112_744 fragme

What is their dimension?

I tried to attach the yambo.in file but it was told that the extension is not allowed.

It is just text, you can just cut&paste it

Best,
m

[tolsen]

Hi

Good point:-) Here it is:

optics # [R OPT] Optics
bse # [R BSK] Bethe Salpeter Equation.
bss # [R BSS] Bethe Salpeter Equation solver
StdoHash= 20 # [IO] Live-timing Hashes
Nelectro= 36.00000 # Electrons number
ElecTemp= 0.000000 eV # Electronic Temperature
OccTresh=0.1000E-4 # Occupation treshold (metallic bands)
More_IO_Path= "." # [IO] Additional I/O directory
Com_Path= "." # [IO] Communication directory
FFTGvecs= 2967 RL # [FFT] Plane-waves
NonPDirs= "none" # [X/BSS] Non periodic chartesian directions (X,Y,Z,XY...)
#KfnQPdb= "E < ../GW/SAVE/ndb.QP" # [EXTQP BSK BSS] Database
KfnQP_N= 1 # [EXTQP BSK BSS] Interpolation neighbours
% KfnQP_E
0.800000 | 1.000000 | 1.000000 | # [EXTQP BSK BSS] E parameters (c/v)
%
LongPath= "none" # [Xd] Longitudinal gauge path
BSresKmod= "xc" # [BSK] Resonant Kernel mode. (`x`;`c`;`d`)
BScplKmod= "none" # [BSK] Coupling Kernel mode. (`x`;`c`;`d`;`u`)
% BSEQptR
1 | 1 | # [BSK] Transferred momenta range
%
% BSEBands
16 | 20 | # [BSK] Bands range
%
BSENGBlk= 111 RL # [BSK] Screened interaction block size
BSENGexx= 2967 RL # [BSK] Exchange components
% BSEEhEny
-1.000000 |-1.000000 | eV # [BSK] Electron-hole energy range
%
BSEClmns=0 # [BSK] Kernel Columns
% BSehWind
100.0000 | 100.0000 | # [BSK] [o/o] E/h coupling pairs energy window
%
BoseCut= 0.10000 # [BOSE] Finite Tel Bose function cutoff
BSSmod= "d" # [BSS] Solvers `h/d/i/t`
% BEnRange
0.00000 | 10.00000 | eV # [BSS] Energy range
%
% BDmRange
0.10000 | 0.10000 | eV # [BSS] Damping range
%
BDmERef= 0.000000 eV # [BSS] Damping energy reference
BEnSteps= 200 # [BSS] Energy steps
% BLongDir
1.000000 | 0.000000 | 0.000000 | # [BSS] [cc] Electric Field
%
WRbsWF # [BSS] Write to disk excitonic the FWs

and the static dielectric function was calculated with the same parameters (100 bands and 111 Gs in the response function).

BR
Thomas

[myrta gruning]

Hallo Thomas

I cannot spot any problem from the input. You use a small number of bands so the size of the fragments should not be that large.

Can you check anyway if there are files exceeding 2G?

Code: Select all

find SAVE/ -size +2G -ls

Which version of Netcdf are you using?

Best,
m

[tolsen]

Hi Myrta

There is no files larger than 2 GB and I am using netcdf 4.0.1.

BR
Thomas

[andrea marini]

Thomas, I see very little possibilities here. The Netcdf error message is not given by Yambo. In mod_IO.F you can find the line

Code: Select all

/src/modules/mod_IO.F:       write (msg,'(2a)') '[NetCDF] ',trim(nf90_strerror(status))

as you can see the error message is composed using the nf90_strerror function that is an internal NetCDF error function. Therefore the problem of "too large file" is internal to Netcdf. From the man page of netcdf you can see that

Code: Select all

4.4 Large File Support
======================
It is possible to write netCDF files that exceed 2 GiByte on platforms
that have "Large File Support" (LFS). Such files are
platform-independent to other LFS platforms, but trying to open them on
an older platform without LFS yields a "file too large" error.

Therefore the only possibility is that you are not using the LFS. Now the point is that at the moment I do not remember the story of the LFS. Yambo must be compiled with the LFS. In addition however, Netcdf itself and/or the platform must support the LFS. Myrta, maybe you remember the details better than me.

Andrea