[FLASH-USERS] Bug fixes for active particles unit

[Chris Daley]

Hi Seyit,

The error you show should be unrelated to the recent patch.

You are seeing a controlled abort in gr_checkGridConsistency
after a Grid consistency check fails.  This subroutine is called
in gr_updateRefinement after a refinement/derefinement happens in
amr_refine_derefine.

Some questions that may be useful:

Do you use a custom gr_markRefineDerefine or
Grid_markRefineDerefine?  Does the error appear with both
Paramesh-4.0 and Paramesh-4dev?  If you manually add an extra
call to gr_checkGridConsistency before amr_refine_derefine in
gr_updateRefinement, does the abort also happen before the
refinement/derefinement?

Chris


Seyit Hocuk wrote:
> Hi Chris,
> 
> After applying your patch, I still get the error as shown below when 
> using active particles. This is a restart from a running simulation 
> which was crashing at the same point. I was hoping that this would now 
> be solved, alas.
> 
> Kind regards,
> Seyit
> 
> 
> refined: total blocks =         3145
> Missing neighbors on PE    2, lb=  8,surrblks=(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(   8,   2, 2)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)
> Missing neighbors on PE    2, lb=  8,surrblks=(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(   8,   2, 2)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  -1,-1)(  -1,  
> -1,-1)
> *** Wrote plotfile to xSHx_hdf5_plt_cnt_0001 ****
> DRIVER_ABORT: missing neighbor!
> 
> 
> 
> 
> 
> Chris Daley wrote:
>> To FLASH users,
>>
>> We have recently become aware of two bugs in the active particles unit
>> of FLASH.  These bugs do not affect tracer particles.
>>
>> 1). There is an error when mapping a portion of a particle's mass to a
>> neighboring block that is at a coarser refinement level.  The bug only
>> happens when a particle is within a zone that is next to a fine-coarse
>> block boundary.  It causes integrated PDEN mass on the grid to differ
>> from the total particle mass.
>>
>> 2). There is an error in the calculated size of a receive buffer which
>> can lead to out-of-bounds accesses.  The bug can only happen in
>> applications which have explicitly included the non-default
>> point-to-point map particles to mesh implementation (i.e. using
>> -unit=Grid/GridParticles/GridParticlesMapToMesh/Paramesh/PttoPt).
>>
>>
>> The bugs can be fixed by applying the patch attached to this email.
>> Navigate to the top level directory of an untouched version of
>> FLASH3.3 and enter the command:
>>
>> patch -p0 < fix_particle_map_jan_2011.diff
>>
>> Additional information about the fixes is provided in the file
>> svn_log_messages.txt.  A copy of all the attachments are at
>> http://www.flash.uchicago.edu/~cdaley/FLASH3.3_patches/fix_particle_map_jan_2011/ 
>>
>>
>> Regards,
>> Chris