YAMBO 3.2.3 rev11 - IOTK error with p2y

[feffeficus]

Hi dear all!

again a problem.

I've compiled yambo-3.2.3 rev11 downloaded from sourceforge on sp6@ cineca.
Everythin was fine during compilation. When I've tried to run p2y,
I found this error regarding IOTK I suppose:

Code: Select all

# FROM IOTK LIBRARY, VERSION 1.1.0development
# UNRECOVERABLE ERROR (ierr=1)

and this is the first of long error message.
I've attached here the file with iotk_error and compilation option, hope that will be useful to understand the problem.
Thanks in advance!

[Daniele Varsano]

Ciao Feffe,

can you try to run p2y in serial???


Daniele

[feffeficus]

Yes:

1) from command line

/sp6/userhpe/hpefr001/codes/yambo-3.2.3-rev.11/bin/p2y -S -N

I got the same errror:

########################################################################################################################
# FROM IOTK LIBRARY, VERSION 1.1.0development
# UNRECOVERABLE ERROR (ierr=1)
# ERROR IN: iotk_getline (iotk_scan.spp:836)
# CVS Revision: 1.7
iostat=-1
# ERROR IN: iotk_scan_tag (iotk_scan.spp:520)
# CVS Revision: 1.7
# ERROR IN: iotk_scan (iotk_scan.spp:714)
# CVS Revision: 1.7
# ERROR IN: iotk_scan_begin (iotk_scan.spp:94)
# CVS Revision: 1.7
########################################################################################################################
ERROR: 0032-150 MPI is not initialized in MPI_Abort, task 0
ERROR: 0031-250 task 0: Terminated

2) through the loadlevel script in serial...I'm queing.....

[feffeficus]

ok. also from by submitting through loadlever I get the same error.

I;'ve found this comment about the same error I have in the pwscf_forum, here the link:

http://www.democritos.it/pipermail/pw_f ... 09753.html

What do you suggest to do so?
ciao.

[Conor Hogan]

Curse you IOTK! *shakes fist*

Can you post your data-file.xml file?

[Conor Hogan]

Well this was a very tricky bug to find...and in fact the bug is in the IOTK library, not in yambo.
The error started in versions later than 3, when we introduced a check of the pseudopotential files to make sure that norm conserving ones were used, and not ultrasoft, so the .UPF files are now being scanned with the IOTK library.
In Feffe's case he had two files:

Code: Select all

H.pz-vbc.UPF
<PP_INFO>
Generated using ld1 code
Author: P. Giannozzi    Generation date: 1990
Info: H LDA 1s1 VonBarth-Car local
...

Si.pz-vbc.UPF
<PP_INFO>
Generated using unknown code
Author: von Barth and Car   Generation date:  < 1984
Info: Si LDA 3s2 3p2 VonBarth-Car, l=2 local

The scan of the H file works fine, the scan of the Si file gives an error: this is due to libiotk getting confused by the "<" in "date: < 1984" !!!

Since none of my efforts to get around this using the IOTK routines worked, I suggest just erase the "<" in the PP_INFO section of your pseudopotentials. Of course this is not read, so it is harmless. The problem is also just limited to that particular file...or files with such a " <" character in the PP_INFO.

Phew!
Conor