Error in reading geometry.xml

Hi, I am encountering errors when I tried to run Openmc with openmc-0.10.0/examples/xml/reflective.

I cloned the Openmc Github with tag = v0.10.0 and build from source.

I have following modules loaded:

Currently Loaded Modules:

  1. intel/2018.3.051 3) cmake/3.12.1-default-gcc

  2. zlib/1.2.11-icc 4) hdf5/1.10.2-parallel-fortran-icc

When I ran Openmc under the ‘reflective’ directory, it reported:

The OpenMC Monte Carlo Code

Copyright | 2011-2018 Massachusetts Institute of Technology

License | License Agreement — OpenMC Documentation

Version | 0.10.0

Date/Time | 2018-09-09 11:37:16

MPI Processes | 1

OpenMP Threads | 44

Reading settings XML file…

Reading cross sections XML file…

Reading materials XML file…

Reading geometry XML file…

forrtl: severe (408): fort: (8): Attempt to fetch from allocatable variable DATA when it is not allocated

Image PC Routine Line Source

openmc 000000000041F200 Unknown Unknown Unknown

libopenmc.so 00002B0166E26E45 algorithm_mp_find 264 algorithm.F90

libopenmc.so 00002B01670A0381 input_xml_mp_read 1879 input_xml.F90

libopenmc.so 00002B016705495D input_xml_mp_read 67 input_xml.F90

libopenmc.so 00002B016704E585 openmc_init 90 initialize.F90

openmc 000000000040AEBA MAIN__ 21 main.F90

openmc 000000000040AD9E Unknown Unknown Unknown

libc-2.17.so 00002B016B551B35 __libc_start_main Unknown Unknown

openmc 000000000040ACA9 Unknown Unknown Unknown

I read some of the code and it should be the problem in the input_xml.F90, Line 1879,

if (find(fill_univ_ids, u % id) == -1)

passes a VectorInt without ‘data’ to

pure function find_vector_int(vec, val) result(index)

But I don’t really know how to fix it…

在 2018年9月9日星期日 UTC+8上午11:46:18,Shen Archon写道:

Hi Shen,

I’ve attached a patch for this problem. You can apply it by running ‘git apply intel_patch’ in the root directory of your OpenMC installation. Let me know if you’re still experiencing problems after using this patch.

Best,
Paul

intel_patch (882 Bytes)

Hi Paul,

Thank you very much for the fix. It works!

However, when running the test suites by ‘run_test.py’, I got ~30 tests failed with ‘AssertionError: Results do not agree.’

I searched online (https://github.com/openmc-dev/openmc/issues/271) and it may be caused by ‘WARNING: Maximum number of sites in fission bank reached. This can result in irreproducible results using different numbers of processes/threads.’

I wonder are these test considered as succeeded or failed as in the github post it says “so I wouldn’t really consider this behavior a bug, per se – it’s just that you end up getting a different answer.”

Best,
Youlin

在 2018年9月10日星期一 UTC+8下午8:54:48,Paul Romano写道:

If you have a machine with a large number of CPU cores, you might want to try to cap the number of OpenMP threads being used. This can be controlled through the OMP_NUM_THREADS environment variable. Try setting it to 2 or 3 and then hopefully the tests will pass.

Best,
Paul

Hi Paul,

I set it to both 2 and 3 and neither of them will let me pass the test suites…

Best,
Youlin

在 2018年9月17日星期一 UTC+8下午11:36:28,Paul Romano写道:

Hi Paul,

I tried again with number of threads = 2,3,22,44 and all of them gave the same result.

The hash value is

3b76b468b9c0e7df6508189b75748a1b7b4f2b37486396749e1a15e536526336f70b60bb207095c39ecbd46822e8c8705ea81184a3c8546e7da09bb905d74637.

Best,

Youlin

在 2018年9月19日星期三 UTC+8下午8:13:42,Shen Archon写道:

Hi Shen,

Which test is the hash from? What cross section data are you using to run the test suite?

Right now some of the tests are only guaranteed to pass if you use the data used by travis. The failure of the test may just mean your cross section data is producing different tallies for some of the tests.

-Isaac

Hi Issac,

The hash is from the test_asymmetric_lattice.py.

I got the cross section data by using:


wget -O nndc_hdf5.tar.xz $(cat openmc/.travis.yml | grep anl.box | awk ‘{print $2}’)
tar xJvf nndc_hdf5.tar.xz


from the online documentation.

Best,
Youlin

在 2018年9月20日星期四 UTC+8下午10:00:12,Isaac Meyer写道: