RESTART calculations :: current_fragment fragments_todo

Various technical topics such as parallelism and efficiency, netCDF problems, the Yambo code structure itself, are posted here.

Moderators: Daniele Varsano, andrea.ferretti, andrea marini, Conor Hogan, myrta gruning

RESTART calculations :: current_fragment fragments_todo

Postby martin.gmitra » Thu Sep 19, 2013 2:54 pm

Dear yambo developers,
I would like to suggest an enlargement of the integer variables for fragments in RESTART files.
src/io/Fragments_Restart.F
line 72: write (81,'(/2x,2(a,i6,1x))') 'Section Completed ',current_fragment,'. To reach ',fragments_todo

In case of mine calculations I have got just:
Section Completed ****** . To reach ******

Is there a way one can figure out current_fragment and fragments_todo done in my stopped
calculations thus I can restart them?

Regards,
Martin Gmitra, PhD
University of Regensburg, Germany
martin.gmitra
 
Posts: 26
Joined: Tue Apr 23, 2013 8:40 pm

Re: RESTART calculations :: current_fragment fragments_todo

Postby Daniele Varsano » Thu Sep 19, 2013 3:21 pm

Dear Martin,
thanks for reporting it,
it looks you have really a lot of fragments!!!
I do not know what in which runlevel you are, in the case you are running with the -S option, then looking at

Code: Select all
ls -ltr ./SAVE


you can see the last fragment wrote in the SAVE file. In any case you will have to change the format of the I/O restart, for instance
i8 insetead if i6, otherwise yambo will not recognize the fragments_to_do number.

Another option, if you are using the netcdf library, you can explore what is inside your databse using the ncdump command, but having a very huge file as I presume, I do not know if it can be deduced easily.

In the worst case, if you doe not succeed to understand the last fragments you arrived, after having modified the format, you can write in the restart file 999999, and the code will restart from there, in this way you recover something.



Hope it helps,

Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 1996
Joined: Tue Mar 17, 2009 2:23 pm

Re: RESTART calculations :: current_fragment fragments_todo

Postby martin.gmitra » Thu Sep 19, 2013 5:36 pm

Dear Daniele,
Thanks for your replay. I am running COHSEX calculations with the options: -b -k hartree -g n -p c
and firing just: yambo -F input -J FOLDER. The database has been prepared from PW DFT using
p2y -S -c. Now I see, that I did not use -S option when running yambo. Does it harms?

Let me list what I do have in ./SAVE folder:
ns.wf_fragments_1_1
...
ns.wf_fragments_75_1

It looks that the fragments in ./SAVE are different from those in ./RESTART file?

As to the I/O format, do you mean the line:
write (81,'(/2x,2(a,i6,1x))') 'Section Completed ',current_fragment,'. To reach ',fragments_todo
to
write (81,'(/2x,2(a,i8,1x))') 'Section Completed ',current_fragment,'. To reach ',fragments_todo
right?
Subsequent reading of the restart file is format-free?

Thank a lot,
Martin Gmitra, PhD
University of Regensburg, Germany
martin.gmitra
 
Posts: 26
Joined: Tue Apr 23, 2013 8:40 pm

Re: RESTART calculations :: current_fragment fragments_todo

Postby Daniele Varsano » Thu Sep 19, 2013 8:21 pm

Dear Martin,

Now I see, that I did not use -S option when running yambo. Does it harms?

It does not harms, but simply yambo does not fragment the files, so what I suggested to look at the last written file will not works.

Let me list what I do have in ./SAVE folder:
ns.wf_fragments_1_1
...
ns.wf_fragments_75_1

It looks that the fragments in ./SAVE are different from those in ./RESTART file?


No, the ns.wf_ are produced by the p2y, so they are fragmented as you used the -S option. I was referring to the ndb.* files, but if you did not use the -S option when running yambo the trick will not apply. Anyway, you do not have any ndb.*? Did not run the initialization first?

As to the I/O format, do you mean the line:
write (81,'(/2x,2(a,i6,1x))') 'Section Completed ',current_fragment,'. To reach ',fragments_todo
to
write (81,'(/2x,2(a,i8,1x))') 'Section Completed ',current_fragment,'. To reach ',fragments_todo
right?
Subsequent reading of the restart file is format-free?


No, I was suggesting to modify the reading part (or both):
Code: Select all
82      read (81,'(/////2x,a,i6)') local_string(:18),io_restart_point(ID)


Anyway, you can try with the ncdump command to look at your last file produced (do not know if the ndb.em1s or ndb.QP as I do not know where the code stopped.If it is not successful may be the safer think to do is to run your calculation from scratch, after having modified the format in both write and read part.

Sorry for not having a more clear solution,

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/
User avatar
Daniele Varsano
 
Posts: 1996
Joined: Tue Mar 17, 2009 2:23 pm

Re: RESTART calculations :: current_fragment fragments_todo

Postby martin.gmitra » Fri Sep 20, 2013 7:39 am

Dear Daniele,

Thanks a lot for your replay. The last modified is ndb.QP. Dumping for a last hint in the ndb.em1s shows that there should be
fine since the last hint is full and corresponds to my last 75th transferred momenta. In ndb.QP I have found filled QP_Sc
variable. From this I can not figure out how many fragments has been done or not. Anyway, I have modified read/write in
source and redo the calculations.

Best,
Martin Gmitra, PhD
University of Regensburg, Germany
martin.gmitra
 
Posts: 26
Joined: Tue Apr 23, 2013 8:40 pm


Return to Technical Issues

Who is online

Users browsing this forum: No registered users and 1 guest

cron