[FLASH-USERS] gr_findAllNeghID

John ZuHone jzuhone at gmail.com
Mon Nov 27 14:41:11 EST 2017


Hi Klaus,

Thanks for your reply, it was helpful.

> 1) Make sure the interface of gr_findAllNeghID is known explicitly where 
> you call it, so you should have something like:
> 
>  use gr_interfaceTypeDecl
>  use gr_interface, ONLY : gr_findAllNeghID

Yes, I already had this in there, but thanks for reminding me to check. 

> 2) You may find the subroutine gr_checkGridState useful
> 
> 3) Subroutine gr_findAllNeghID is also applied in the code that couples 
> FLASH to HYPRE, so maybe you can find a useful hint for use in 
> source/Grid/GridSolvers/HYPRE/paramesh/gr_hypreSetupGrid.F90 .

I had a look at these--the use of gr_findAllNeghID seems pretty identical to what I am doing, so I’m not sure what’s going on here. 

> 4) How many blocks and procs do you have?
> Are the block and proc numbers that occur in the following lines 
> reasonable at all?
> 
>> . My block is:         542 and proc is:         540 
>> . My block is:        1082 and proc is:        1080 
> 
> It looks suspicious that block and proc numbers are so close...

These do look odd--I hadn’t noticed that. The simulation does run fine without my special module turned on. 

> Anyway, does this problem occur even on 1 proc?
> Does it occur always, or only starting at a certain number of blocks, and/or 
> in particular grid configurations?

This did not happen in a slightly different setup where all the blocks were on the same refinement level, and it was also multiple processors--in this case the grid is highly refined. 

> Also, what exactly is the information on neighboring blocks that you want 
> to access? I am asking because there may already be different ways to get 
> at the information, as for example Grid_getBlkNeighLevels for neighboring 
> refinement levels.

Basically, I want to figure out the block number and processor of every leaf block that is touching my leaf block, whether on a face, edge, or corner. It seemed like this was the most straightforward way to do it. 

Best,

John




More information about the flash-users mailing list