em1s too time consuming

[Stephan]

Hi Daniele,
thanks for the reply. In the attachment you can find the results of the command ls -ltr ./SAVE

Thanks and Regards
Stephan

[Daniele Varsano]

Dear Stephan,
it looks the screening ndb.em1s is not present in your SAVE directory.
Have you moved it somewhere, or you used the -J redirection command in the em1s run?
Quite strange as in the previous report it was read.

Best,

Daniele

[Stephan]

Sorry these were the wrong data. This should be the correct one.
The amount of data is far too high to upload. Thats why I had to split it up in 14 files a 1MB.
Three of these files are packed together.

Thanks and regards

Stephan

[Stephan]

and the missing one ...

[Daniele Varsano]

Dear Stephan
ok, first of all let me give you an *important* advice. You have a too much fragmented BS kernel and this can gives you huge problem to your file system.
We still have to organize this part in order to avoid that. In the meanwhile I strongly suggest you to insert in your input file this variable:

Code: Select all

DBsIOoff= "BS" 

In this way you will not write the BSE. If you really need it to be stored you can use:

Code: Select all

DBsFRAGpm= "-K -BS" 

Please note the "-" sign.

Next: it looks you have all the needed file, and looks also that you succeed to run a BSE calculation the first time, next something odd has happened.

I suggest you to remove all the previous BS file: ./SAVE/ndb.BS_Q1_CPU_*
next build up an input file by also adding the "-b" option. This should read the ndb.em1s file you previously calculated. In any case check that the
variable (bands in X, Gblk etc.) you find in the input are the *same* of the em1s run you did before. And run again the BS.
Yambo should skip the em1s calculation, read the ndb,em1s files and start to build the BSE kernel.
If it does not work send again the report and standard outputs and we will try to understand what is happening.
Best,
Daniele

[Stephan]

Thank you very much I try what you suggested

[Stephan]

Hello Daniele,
I tried what you suggested and modified the input file.
Again the calculation did not come to an end after three days with 16 procs (maximum capacity I can use).
Can you help me to come to a result?

Thanks and Regards

Stephan

[Stephan]

Here are the remaining log-files and the input-file

Best regards

[Daniele Varsano]

Dear Stephan,
I can see that now the screening is correclty read.
What I suggest you to do is:
1) remove any ndb.BS* file in your SAVE directory if any
2) remove DBsIOoff= "BS" form input and add
DBsFRAGpm= "-K -BS"
in this way you can try to do a restart.
3) Use the parallelization over k points:
BS_CPU= "16 1 1"

Best,
Daniele

[Stephan]

Thank you very much. Let's see if it will work..