many warning with parallel run

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
weitong
Posts: 26
Joined: Mon Sep 27, 2010 5:16 am

many warning with parallel run

Post by weitong » Sat Mar 29, 2014 8:46 am

I am running a test with abinit 7.6.2.
With mpirun, lots of warnings appear in the log file, making it larger than 4MB. For uploading the log file, I have to deleted some of the contents. Please help.
PS: abinit was make with mpich-3.1.

Here are some examples of the warnings:

Code: Select all

--- !WARNING
message: |
    the minimization is stopped for this block
src_file: lobpcgwf.F90
src_line: 855
...
--- !WARNING
message: |
    abi_xpotrf, info=1
src_file: abi_xorthonormalize.f90
src_line: 115
...


--- !WARNING
message: |
    Problem in abi_xheev, info=   2
src_file: abi_xheev.f90
src_line: 184

--- !WARNING
message: |
    Problem in abi_xhegv, info=   6
src_file: abi_xhegv.f90
src_line: 187
...


--- !WARNING
message: |
    abi_xpotrf, info=1
src_file: abi_xorthonormalize.f90
src_line: 115

Attachments
test1.log
(914.49 KiB) Downloaded 287 times
test1.out
(35.87 KiB) Downloaded 279 times

User avatar
gmatteo
Posts: 291
Joined: Sun Aug 16, 2009 5:40 pm

Re: many warning with parallel run

Post by gmatteo » Tue Apr 01, 2014 10:17 pm

The eigensolver went bananas! Did you look at the results in the output file?

iter Etot(hartree) deltaE(h) residm vres2
ETOT 1 -96.680943421047 -9.668E+01 4.956E-03 2.681E+03
ETOT 2 -97.887586727741 -1.207E+00 3.973E-04 2.248E+03
...

ETOT 17 -100.16163378424 1.869E-06 1.394E-13 1.971E-10
ETOT 18 NaN NaN 2.435E-10 NaN
ETOT 19 NaN NaN 0.000E+00 NaN

Can you reproduce the problem if you use paral_kgb=0?
(I assume that the pseudos are OK, and the physics of the system is correctly described)

weitong
Posts: 26
Joined: Mon Sep 27, 2010 5:16 am

Re: many warning with parallel run

Post by weitong » Mon Apr 07, 2014 3:20 am

Thanks.
I checked output file, many NaN. These warnings happen with abinit7.6.2+mpich3+Intel compiler(ifort 12.1.5 or ifort 13.1)
I also tested mpirun the script with abinit7.6.2+openmpi1.6+GNU compiler, there is no warnings.

I am newer to Abinit. How to use paral_kgb=0

User avatar
gmatteo
Posts: 291
Joined: Sun Aug 16, 2009 5:40 pm

Re: many warning with parallel run

Post by gmatteo » Mon Apr 07, 2014 3:27 am

Set paral_kgb 0
in the input file

weitong
Posts: 26
Joined: Mon Sep 27, 2010 5:16 am

Re: many warning with parallel run

Post by weitong » Tue Apr 08, 2014 12:39 pm

Thanks. I added a line "paral_kgb 0" into the input file. Then it works fine without warnings anymore.

Could you please explain a little more about why this happens, and how does "paral_kgb 0" affect the computing results?

Best,
WT

Locked