From NWChem
Viewed 2308 times, With a total of 7 Posts
|
|
|
Clicked A Few Times
Threads 1
Posts 9
|
|
| 12:27:56 AM PDT - Mon, Jun 13th 2011 |
|
Hi, we have been having problems building and running NWChem with the Intel compilers (version 6.0, as well as 5.1.1 and 5.1). There is usually no problem compiling NWChem, and we can usually run a small calculation such as the one below:
echo
start h2o
title "H2O"
charge 0
geometry
O 0.000000 0.000000 0.116577
H 0.000000 0.769424 -0.466309
H 0.000000 -0.769424 -0.466309
end
basis spherical
H library aug-cc-pVDZ
O library aug-cc-pVDZ
end
dft
xc becke88 perdew86
mult 1
end
task dft
However, as soon as we try to run a larger calculation, the job will fail. There isn't much information to see what's going on, but the error message can be:
mpiexec: Warning: task 0 died with signal 11 (Segmentation fault).
or
rank 0 in job 5 computea-01_34827 caused collective abort of all ranks
exit status of rank 0: killed by signal 9
One is with MPICH and the other with Intel MPI. We used Intel compilers version 11.1 on RHEL/CentOS 5.4 or 5.5.
If more information is needed to find out what's going, I will be glad to provide them.
Thank you very much!
Shenggang.
|
|
|
|
|
|
Just Got Here
Threads 0
Posts 4
|
|
| 3:19:22 PM PDT - Mon, Jun 13th 2011 |
|
Re: NWChem 6.0 with Intel Compilers
|
Hi Shenggang,
Can you send me the full output of the failed calculation, please? Usually there is something in the output that suggests what the cause of the trouble is. However, it may not always appear in the most obvious place. So I would like to be able to take a look for myself.
Best wishes,
Huub van Dam
|
|
|
|
|
|
Clicked A Few Times
Threads 1
Posts 9
|
|
|
|
-
Bert Forum:Admin, Forum:Mod, NWChemDeveloper, bureaucrat, sysop
|
|
|
Forum Regular
Threads 2
Posts 285
|
|
| 2:02:07 PM PDT - Mon, Jul 18th 2011 |
|
This is the same issue as in another posting here. It seems like a compiler issue with Intel 11, as it works when compiled with GNU (gfortran I assume). Asked Shenggang if there is another version of the Intel compiler available.
Bert
|
|
|
|
|
|
|
| 5:17:01 PM PDT - Fri, Jul 22nd 2011 |
|
NWChem 6.0 with Intel compilers
|
I think Dr. van Dam is referring to my numerous postings concerning the trouble I had using Intel compilers.
My solution was to use gfortran which is a part of gcc. Note that I am using the most recent version 4.6. I also had run into memory problems using MPICH1. Simply using MPICH2 fixed those problems.
Let me know if this helps.
Also, be sure to run the script that automatically sets your memory to the correct value. I'll bet your system has more than 400MB per core.
|
|
|
|
|
|
|
| 1:33:42 AM PDT - Thu, Aug 11th 2011 |
|
On one of the system, the MPI coming with the system was not built with gfortran but g77. On the other system, I am using the Intel MPI library, so I am not sure I can use gfortran in place of ifort.
It could be due to a problem with MPICH1, but the Intel MPI library is MPICH2, and the problem persists.
Memory shouldn't be a problem, as this DFT job actually use very little memory.
Quote: Jul 23rd 12:17 amI think Dr. van Dam is referring to my numerous postings concerning the trouble I had using Intel compilers.
My solution was to use gfortran which is a part of gcc. Note that I am using the most recent version 4.6. I also had run into memory problems using MPICH1. Simply using MPICH2 fixed those problems.
Let me know if this helps.
Also, be sure to run the script that automatically sets your memory to the correct value. I'll bet your system has more than 400MB per core. |
|
|
-
Bert Forum:Admin, Forum:Mod, NWChemDeveloper, bureaucrat, sysop
|
|
|
Forum Regular
Threads 2
Posts 285
|
|
| 2:06:40 PM PDT - Tue, Aug 16th 2011 |
|
Shenggangli,
Please contact me directly at bert.dejong@pnnl.gov so we can resolve your issues once and for all. You're compiling with all kinds of NWChem versions, all kinds of compilers and libraries, it is unclear what is going on anymore. Let's start with ONE platform, and ONE environment and go from there. We'll post the resolution on here once we're done.
Bert
|
|
|
|
|
|
Clicked A Few Times
Threads 1
Posts 9
|
|
| 12:46:44 AM PDT - Fri, Aug 19th 2011 |
|
Bert,
all right, I will write you by email. Thanks,
Shenggang.
|
|
|
AWC's:
2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC
