parallelism

[zaabar foudil]

Dear Daniele
I have the possibility to run my GW calculations and BSE in our cluster with 4 (node) * 16 (ntasks-per-node) in our cluster, to minimize the computation time and the problem of memory, how can I launch them using the flague contained in the input files for GW and BSE
BS_CPU = "" # [PARALLEL] CPUs for each role
BS_ROLEs = "k eh t" # [PARALLEL] CPUs roles (k, eh, t)
...........
you find the attached files for my previous calculations : GW, bse l_opt, yambo.in .....,
Note; I ran these calculations with default option " 4 node * 4 tasks" due to the memory problem which took me more than 18 days for each calculation.
Here are some details of our cluster:
#######################################################################
processor : 15
vendor_id : GenuineIntel
cpu family : 6
model : 45
model name : Intel(R) Xeon(R) CPU E5-2660 0 @ 2.20GHz
stepping : 7
microcode : 0x710
cpu MHz : 1199.945
cache size : 20480 KB
physical id : 1
siblings : 8
core id : 7
cpu cores : 8
apicid : 46
initial apicid : 46
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes xsave avx lahf_lm ida arat epb pln pts dtherm tpr_shadow vnmi flexpriority ept vpid xsaveopt
bogomips : 4404.13
clflush size : 64
cache_alignment : 64
address sizes : 46 bits physical, 48 bits virtual
power management:
################################################################################################

my best regards
zaabar foudil
university of bejaia , algeria

[Daniele Varsano]

Dear zaabar,
here you can find a tutorial on the parallelization strategies for GW implemented in Yambo.
http://www.yambo-code.org/wiki/index.ph ... strategies

For the Bethe Salpeter (and also other runlevels) parallelism you can find information in the preprint of this paper
describing the new features implemented in Yambo.

https://arxiv.org/pdf/1902.03837.pdf

Note that you can also use default parallelization strategy using the variable:

Code: Select all

PAR_def_mode= "balanced"       # [PARALLEL] Default distribution mode ("balanced"/"memory"/"workload")

"memory" is meant to reduce the memory load for each core.

Best,
Daniele

[zaabar foudil]

Dear Daniele
I thank you for your guidance and your valuable advice.
I have two other questions:
1 - how to fix the following warning:
[WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file
2- I have successfully installed the latest version of yambo-4.3.2 with the intel compiler, it works well, but it does not show me the folder / LOG.

my best regards
zaabar foudil
university of bejaia , algeria

[Daniele Varsano]

Dear zaabar,

1 - how to fix the following warning:
[WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file

No need to fix, it is just a warning, if you think (or better you have verified) that the term [x,Vnl] can be neglected in the dipole calculation you can speed up the calculation renaming the ns.kb_pp in the SAVE directory.

2- I have successfully installed the latest version of yambo-4.3.2 with the intel compiler, it works well, but it does not show me the folder / LOG.

This means you are running with just 1 core, if this is not the case it means that the parallel compilation of yambo was not successful.

Best,
Daniele

[zaabar foudil]

Dear Daniele
I tried to perform the calculation GW with the latest version (yambo-4.3.2), it works well but I did not get the values of the GW.
here are all input and output files
thank you in advance

zaabar foudil
university of bejaia , algeria

[Daniele Varsano]

Dear zaabar,
something went wrong in the calculation of the local potential matrix elements <Vxc>.

Code: Select all

[xc] E_xc :   0.000000 [Ha]
      E_xc :   0.000000 [Ry]

and you can see all the elements are NaN--> Not a Number

Code: Select all

 <43|HF|43> = -18.00069 0.354E-09 <43|DFT|43> =       NaN       NaN

Now it is hard to understand what went wrong.
I suggest you
1. delete the ./SAVE/ndb.HF_and_locXC file
2. check you have the latest release
3. repeat the calculation (only the HF part, yambo -x -r) by using more plane waves:

Code: Select all

FFTGvecs=  16893
EXXRLvcs= 16893

If the problem persists please post here you QE scf/nscf file and pseudopotentials and we try to reproduce your problem.

Best,

Daniele

[zaabar foudil]

Dear Daniele

i did what did you told me about, step by step, but the problem remains unchanged, here are the scf , ncsf and pseudopotential files
Note : the same script of the scf and ncsf work well in other cluster

best regard


zaabar foudil
university of bejaia , algeria

[zaabar foudil]

Dear Daniele

here are the other pseudo files



best regard


zaabar foudil
university of bejaia , algeria

[Daniele Varsano]

Dear zaabar,
ok, I will take a look as soon as I find the time.
Anyway, in your previous message you are saying that yambo compiled in a different cluster does not suffer of this problem?

Daniele

[zaabar foudil]

Dear Daniele
to see if the origin of the previous problem comes from Q-e or yambo, I transferred a folder "prefix.SAVE" of my laptop to the current cluster, the problem remained the same. I think the problem comes from yambo code
Best,
zaabar foudil
university of bejaia , algeria