Questions on external source file reading

User Support
1

Hello everyone,

I’ve got some questions on OpenMC external source file reading. While I run a fixed source problem, I get an “invalid memory reference” error, the external source is given in source.h5 file.

Inputfiles.zip (266 KB)

2

Hi Shawn,

As I noted in a response to your other post, right now fixed source calculations using a source file are broken. I will look into getting this fixed for the next version of OpenMC.

Best regards,
Paul

3

Hi Paul，

I found in the subroutine finalize_generation in simulation.f90, you write:
! Write generation output
if (master .and. current_gen /= gen_per_batch) call print_generation()
elseif (run_mode == MODE_FIXEDSOURCE) then
! For fixed-source mode, we need to sample the external source
do i = 1, work
call set_particle_seed(overall_gen*n_particles + work_index(rank) + i)
call sample_external_source(source_bank(i))
end do
The problems may be out here, you didn’t separate the source file reading away from external source sampling. After commenting out the do cycle here, OpenMC could run in serial calculation, but problem still exist in parallel calculation, I don’t know why this happens. Besides, the results given by fixed source calculation using FLUKA-OpenMC vary too much from those given directly by MCNPX. Is there any validations of fixed source calculation of OpenMC? Or is it suitable to calculate an ADS problem by coupling FLUKA and OpenMC?
Looking forward to your response.
Best wishes,
Shawn
在 2017年1月4日星期三 UTC+8下午9:47:41，Paul Romano写道：

4

Hi Shawn,

We discovered and fixed the issue in finalize_generation just the other day. Could you describe the problem you are seeing when running in parallel? Also, it’s hard to give any specific guidance regarding your coupled FLUKA-OpenMC calculation compared to MCNPX without knowing more about what you’re doing. How big are the differences that you are seeing?

Best,
Paul

5

Hi Paul,
The OECD-ADS benchmark flux and kappa-fission calculations fit well(the relative difference is about 2%), but for our ADS model, the results are 50% smaller than those given by MCNPX. I think something might be wrong in my input files and I’m working on it.
Best wishes,
Shawn

在 2017年1月10日星期二 UTC+8上午2:09:33，Paul Romano写道：

6

Now my OpenMC could run in parallel mode. That problem occurs because I forgot the “make install” step.

在 2017年1月11日星期三 UTC+8下午8:19:23，Shawn写道：