[FLASH-USERS] Fatal error in MPI_Comm_create when I run Flash on Mac os x

[Klaus Weide]

On Tue, 9 Jul 2013, Reem Alraddadi wrote:

> I am running laser slab example after I modified for three materials, and I
> am using the following command :
> 
> /setup -auto Slab3 -2d +pm4dev  -nxb=16  -nyb=16  species=cham,targ,foil
> +mtmmmt +mgd  mgd_meshgroups=6 -parfile=example.par +3t  -maxblocks=3000

Reem,

You may also want to add +uhd, or better use +uhd3t instead of just +3t,
tu use the unsplit Hydro solver. (I assume that your current setup is 
using the split Hydro solver, but I can't be sure - it depends on what
REQUIRES and REQUESTS are listed in the Slab3 simulation's Config file.)

> I increased the maxblocks to 3000 as I used 6 level of AMR.In addition,
> Yes, I am using hypre-2.9.0b with FLASH4.0.1. I will change it to
> hypre-2.80b. However, the library that I used together with hypre to run
> FLASH4.0.1 are:
> 1-MPICH-3.0.4
> 2.hdf5-1.8.9
> 
> 
> However, may I ask different question here. Is increasing maxblocks to 3000
> is fine to run with 4 GD memory of Mac? Do I need more memory in order to
> increase the resolution here?

You are in the best position to find that out, for the simuation you are 
interested in and for the the system and setup options you want to use.
You WILL notice when you run out of memory!

You can estimate how much memory you will need just to store your 
solution, i.e., the UNK array with 'maxblocks' blocks, on *each* MPI 
process:

   (16+8) * (16+8) * NUNK_VARS * maxblocks   double precision numbers

There is also significant memory required for some other purposes,
esp. scratch arrays when using the unsplit Hydro solver.

Klaus