NC_MAX_VARS error without yambo -s

[Yadong Wei]

Dear Yambo Users:
I am calculating excitonic optical absorption of an 2D system with BSE. Just following the steps of the tutorial

http://www.yambo-code.org/wiki/index.ph ... al_systems

However I got errors in the main loop in BSE with the message so many times:

Code: Select all

 
<14m-06s> P0003: Kernel |########                                | [020%] 13m-35s(E) 01h-07m-55s(X)
<16m-09s> P0003: Kernel |#########                               | [022%] 15m-37s(E) 01h-09m-27s(X)
P0003: [ERROR] STOP signal received while in :[07.04.03] Kernel loop
P0003: [ERROR][NetCDF] NetCDF: NC_MAX_VARS exceeded

From the topic

viewtopic.php?f=10&t=349

one of the solution is adding -S option
but in yambo-4.1.2(what I am using) it is invalid option, here is the help page (yambo -H page)

Code: Select all

 Tool: yambo 4.1.2 rev.14024
 Description: A shiny pot of fun and happiness [C.D.Hogan] 

 -h		:Short Help
 -H		:Long Help
 -J <opt>	:Job string identifier
 -V <opt>	:Input file verbosity
		 [opt=RL,kpt,sc,qp,io,gen,resp,all,par]
 -F <opt>	:Input file
 -I <opt>	:Core I/O directory
 -O <opt>	:Additional I/O directory
 -C <opt>	:Communications I/O directory
 -D		:DataBases properties
 -W <opt>	:Wall Time limitation (1d2h30m format)
 -Q		:Don't launch the text editor
 -M		:Switch-off MPI support (serial run)
 -i		:Initialization
 -o <opt>	:Optics [opt=(c)hi is (G)-space / (b)se is (eh)-space ]
 -k <opt>	:Kernel [opt=hartree/alda/lrc/hf/sex]
		        (hf/sex only eh-space; lrc only G-space)
 -y <opt>	:BSE solver [opt=h/d/(p/f)i]
		        (h)aydock/(d)iagonalization/(i)nversion
 -r		:Coulomb potential
 -x		:Hartree-Fock Self-energy and local XC
 -d		:Dynamical Inverse Dielectric Matrix
 -b		:Static Inverse Dielectric Matrix
 -p <opt>	:GW approximations [opt=(p)PA/(c)HOSEX]
 -g <opt>	:Dyson Equation solver
		 [opt=(n)ewton/(s)ecant/(g)reen]
 -l		:GoWo Quasiparticle lifetimes
 -a		:ACFDT Total Energy

By	YAMBO developers group
	http://www.yambo-code.org

So is there a way to get this option -S in this release package? Or any other approach to solve the netCDF problem?

my lumen -D message shows below

Code: Select all

[RD./SAVE//ns.db1]------------------------------------------
 Bands                           : 30
 K-points                        :  729
 G-vectors             [RL space]:  68033
 Components       [wavefunctions]:  8559
 Symmetries             [spatial]: 1
 Spinor components               : 2
 Spin polarizations              : 1
 Temperature                 [ev]:  0.02585
 Electrons                       : 18.00000
 WF G-vectors                    : 10904
 Max atoms/species               : 1
 No. of atom species             : 3
 Magnetic symmetries             : no
- S/N 006831 -------------------------- v.01.00.00 r.00120 -
[RD./SAVE//ns.wf]-------------------------------------------
 Fragmentation                    :yes
 Bands in each block             : 30
 Blocks                          : 1
- S/N 006831 -------------------------- v.01.00.00 r.00000 -
[RD./SAVE//ndb.kindx]---------------------------------------
 Fragmentation                    :no
 Polarization last K   :  729
 QP states             :    1   729
 X grid is uniform     :yes
 BS scattering         :yes
 COLL scattering       :no
- S/N 006831 -------------------------- v.01.00.00 r.00000 -
[RD./SAVE//ndb.dip_iR_and_P]--------------------------------
 Brillouin Zone Q/K grids (IBZ/BZ):  729   729   729   729
 RL vectors                   (WF): 2187
 Fragmentation                    :yes
 Electronic Temperature        [K]: 300.0000
 Bosonic    Temperature        [K]: 300.0000
 X band range           :  1  22
 X band range limits    : 22   1
 X e/h energy range [ev]:-1.000000 -1.000000
 RL vectors in the sum  : 2187
 [r,Vnl] included       :yes
 Using shifted grids    :no
 Using covariant dipoles:yes
 Using G-space approach :no
 Using R-space approach :no
 Direct v evaluation    :no
 Field momentum norm    :0.1000E-4
 Wavefunctions          :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
- S/N 006831 -------------------------- v.01.00.00 r.00000 -
[RD./SAVE//ns.kb_pp_pwscf]----------------------------------
 Fragmentation                    :yes
- S/N 006831 -------------------------- v.01.00.00 r.00000 -
[RD./SAVE//ndb.em1s]----------------------------------------
 Brillouin Zone Q/K grids (IBZ/BZ):  729   729   729   729
 RL vectors                   (WF): 2187
 Coulomb cutoff potential         :none
 Fragmentation                    :yes
 Electronic Temperature        [K]: 300.0000
 Bosonic    Temperature        [K]: 300.0000
 Static diel. fun.  energies      :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
                    wavefunctions :Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
 Global Gauge                     :length
 X matrix size          : 203
 X band range           :  1  30
 X e/h energy range [ev]:-1.000000 -1.000000
 X Time ordering        :R
 X xc-Kernel            :none
 X Drude frequency      : 0.00      0.00
 X poles           [o/o]: 100.0000
 RL vectors in the sum  : 2187
 [r,Vnl] included       :yes
 Longitudinal Gauge     :yes
 Field direction        :0.1000E-4  0.000     0.000
 BZ energy Double Grid  :no
 BZ energy DbGd points  :0
 :: Current Q-pt index     : 1
 :: X energy range     [ev]: 0.00      0.00
 :: X damping range    [ev]:  0.10000   0.10000
 :: Number of frequencies  : 1
- S/N 006831 -------------------------- v.01.00.00 r.00000 

-


Thank you sincerely

Yadong Wei
Ph. D student
Department of Physics
Harbin Institute of Technology, Harbin, China

[Daniele Varsano]

Dear Yadong,
my first and strong suggestion is to switch to a more recent version of yambo, many bugs have been fixed in the 4.2 release.
Next, you can try the following workarounds:
You can add in your input the following variables:

Code: Select all

 DBsIOoff= "BS"  

in this way yambo will not write on disk the BSE matrix, thus avoiding the netcdf problem.
Alternatively you can add:

Code: Select all

DBsFRAGpm= "+BS"

this will tell yambo to fragment the BS databases. Note that with this option there is the risk to have a lot of small databases, this also should solve the netcdf problem, at the risk to have too much files that could cause problem to the filesystem.
The first option may is more favourable.

Best,
Daniele

[Yadong Wei]

Dear Daniele:
Thank you for your fast replying. The task works well with DBsIOoff= "BS" . In another way fragmenting the BS database, the file size was set at 4096 bytes. Is there a parameter in the yambo file to alter this value? I think twice or three times of this size can avoid file system error more efficiently, without losing BSE matrix in the disk.
And there is another question: I just installed yambo-4.2.2 but still no -S option shows in the yambo -H code. It that option cancelled, or which configuration should I alter when compiling?

Code: Select all

 
 This is yambo 4.2.2 rev.110
 A shiny pot of fun and happiness [C.D.Hogan] 

 -h		:Short Help
 -H		:Long Help
 -J <opt>	:Job string identifier
 -V <opt>	:Input file verbosity
		 [opt=RL,kpt,sc,qp,io,gen,resp,all,par]
 -F <opt>	:Input file
 -I <opt>	:Core I/O directory
 -O <opt>	:Additional I/O directory
 -C <opt>	:Communications I/O directory
 -D		:DataBases properties
 -W <opt>	:Wall Time limitation (1d2h30m format)
 -Q		:Don't launch the text editor
 -E <opt>	:Environment Parallel Variables file
 -M		:Switch-off MPI support (serial run)
 -i		:Initialization
 -o <opt>	:Optics [opt=(c)hi is (G)-space / (b)se is (eh)-space ]
 -k <opt>	:Kernel [opt=hartree/alda/lrc/hf/sex]
		        (hf/sex only eh-space; lrc only G-space)
 -y <opt>	:BSE solver [opt=h/d/(p/f)i]
		        (h)aydock/(d)iagonalization/(i)nversion
 -r		:Coulomb potential
 -x		:Hartree-Fock Self-energy and local XC
 -d		:Dynamical Inverse Dielectric Matrix
 -b		:Static Inverse Dielectric Matrix
 -p <opt>	:GW approximations [opt=(p)PA/(c)HOSEX]
 -g <opt>	:Dyson Equation solver
		 [opt=(n)ewton/(s)ecant/(g)reen]
 -l		:GoWo Quasiparticle lifetimes
 -a		:ACFDT Total Energy

 YAMBO developers group (http://www.yambo-code.org)

My configuration code is

Code: Select all

./configure --with-iotk-path="/public/software/qe-5.4/iotk/" FC=ifort --enable-netcdf-LFS 

and configuration config.log shows in the attachment.

Thank you sincerely
Yadong Wei
Ph. D student
Department of Physics
Harbin Institute of Technology, Harbin, China

[Daniele Varsano]

Dear Yadong,
the -S option is deprecated, fragmentation is now handled using the DBsFRAGpm variable.
Do you still have the same problem also using the 4.2 version?
Unfortunately there is no possibility to choose the size of the fragments.
Anyway, unless you have millions of fragments you can sill keep that fragmentation on.
Another possiiblity is to raise the number of CPU as even without fragmenting the databases, the BS matrix will be splitted among CPUs.

Best,
Daniele