Lifetimes

[amolina]

Hello,

it happens to me very often when I calculate the lifetimes (yambo -l -V all) this error:

[ERROR] STOP signal received while in :[04.03] Dynamical Dielectric Matrix
[ERROR]Incomplete Parallel Index Filling

I have read previous posts but still I don't understand very well what is going on...

Any help is welcome...

Cheers,

Alejandro.

[Daniele Varsano]

Dear Alejandro,
can you please post what release/version you are running, how mani cpu you are using and post input/report files?

Best,
Daniele

[amolina]

Sure. Here you have the l_ and r_ file together with the input file.

This job was running in a 16 cores node.

Cheers,

Alejandro.

[Davide Sangalli]

Ciao,
I'm not sure it has any influence, but why do you have

Code: Select all

%QPkrange                    # [GW] QP generalized Kpoint/Band indices
  1| 61| 13| 26|
%

and

Code: Select all

%QPerange                    # [GW] QP generalized Kpoint/Energy indices
  1| 19| 0.0|-1.0|
%

I think you should set 61 also in QPerange as well. However I do not think this would solve the problem ...

Does it happen often / random / on specific runs ?
As always the best thing is if you can reproduce the problem on a small run ...

Davide

[amolina]

Ciao Davide,

I have been able to do the calculations in a 12x12 grid, but always I try in bigger one it crashes with this error. It doesn't matter how many processors I use, in parallel or serial. I will do some more test and I will post them here.

Cheers,

Alejandro.

[Davide Sangalli]

Ok. I think 14x14 is still reasonable to make tests.
If you can send me the pwscf or abint input file with psps and then the yambo input file which generates the error I'll check what is happenning.

D.

[amolina]

One fast question. Can I increase the variable NGsBlkXd? It seems is here the problem.

[Davide Sangalli]

Yes you can. But not of an arbitrary number.
Yambo change it automatically to the closest numer which gives you "a closed shell in G-space".

Some of these numbers are in the report file.
For example in the one you attached

Code: Select all

  [02.03] RL shells
  =================

  Shells, format: [S#] G_RL(mHa)

   [S1932]:36399(0.7996E+5) [S1931]:36387(0.7993E+5) [S1930]:36363(0.7991E+5) [S1929]:36339(0.7987E+5)
   [S1928]:36315(0.7976E+5) [S1927]:36267(0.7974E+5) [S1926]:36243(0.7973E+5) [S1925]:36219(0.7971E+5)
   [S1924]:36195(0.7971E+5) [S1923]:36183(0.7970E+5) [S1922]:36171(0.7970E+5) [S1921]:36147(0.7969E+5)
   [S1920]:36123(0.7968E+5) [S1919]:36099(0.7965E+5) [S1918]:36075(0.7960E+5) [S1917]:36051(0.7958E+5)
   [S1916]:36027(0.7945E+5) [S1915]:36015(0.7939E+5) [S1914]:36003(0.7932E+5) [S1913]:35991(0.7930E+5)
   [S1912]:35967(0.7928E+5) [S1911]:35943(0.7926E+5) [S1910]:35919(0.7926E+5) [S1909]:35907(0.7916E+5)
   [S1908]:35895(0.7916E+5) [S1907]:35883(0.7915E+5) [S1906]:35859(0.7912E+5) [S1905]:35835(0.7907E+5)
   ...
   [S12]:57( 892.3997) [S11]:45( 811.8315) [S10]:33( 789.5682) [S9]:31( 763.4907)
   [S8]:19( 747.3770) [S7]:13( 580.0910) [S6]:11( 402.8409) [S5]:9( 257.8182)
   [S4]:7( 145.0228) [S3]:5( 64.45455) [S2]:3( 16.11364) [S1]:1( 0.000000)

So you can set the varaible to "45 RL" or to "57 RL" (shells 11 and 12 respectively) . If you try to set "50 RL", yambo will go for the closest. In this case "45 RL". Not all numbers are reported but you do not need to care about, just when you increase the variable try to increase it of a reasonable number.

Also I suggest to use energy in input. It is more physical in non 3D systems. You can chose, instead of "45 RL", "810 mHa" (yambo will round it to "~812 mHa") or also you can use other energy units (for example "eV").
So for example you can increase it of 100 mHa each time or of 5 eV (yambo reads it as an integer number).

Does this solve the issue ?

Davide

D.

[amolina]

Well, it works partially, not very consistenly. It seems if I increases the number, yambo calculate the X but not the lifetimes...