[FLASH-USERS] hangup with laser2Din3D

[Klaus Weide]

On Thu, 21 Apr 2016, Mario Manuel wrote:

> Hello FLASH users,
> 
> I am running a FLASH4.2.2 simulation that is a combination of the 
> LaserSlab and FieldLoop example simulations. It is a laser-matter 
> interaction (multi-material, 3T, tabulated EOSs) with MHD as well.  I 
> recently implemented a more complex laser setup with multiple beams and 
> wanted to use the ed_laser3Din2D = .true. for a more accurate ray 
> tracing in the R-Z geometry that I am using.  After fixing all the 
> initialization errors, I got it to run, but it never gets in to the 
> first step. The simulation hangs up, but doesn't throw an error. ...

Mario, here are some ideas:

1. Use the latest released version (FLASH 4.3), if possible, especially 
if some debugging is necessary.  

2. Try without threading. I.e., re-setup without threading (most 
thorough). (As a quicker test, you may also try to sent the 
environment variable OMP_NUM_THREADS to 1, and/or set the various thread*
runtime parameters that are configured into your simulation to FALSE.

2a. When using threading, use something like

     ulimit -s 65536

 and/or

     export OMP_STACKSIZE=16M

(maybe with larger numbers) in the shell environment before running FLASH 
(and do whatever your system requires so that the FLASH processes will 
inherit these settings). This tries to ensure that threads have enough 
memory for their stacks.

3. I don't think you were running out of memory overall. Actually you
have fewer blocks than MPI tasks, so some processes get to work on an 
empty block list.  This could cause some problems (even though it should 
not), especially in connection with threading.

4. Use the lens-related runtime parameters correctly, as Norbert already 
pointed out.

HTH,

Klaus