[mpas-developers] Mpas Bug Fix

[Michael Duda]

Hi, Doug.

Excellent work by you and the LLNL folks. After looking at the
code in module_block_decomp.F, I'd agree that line 112 isn't
correct, and the fix you provide below is precisely what is
needed. Without the fix, I'm not able to run an MPI-compiled MPAS
executable with a single task, while I can get a correct run
through with your fix applied. Unless anyone else has any
suggestions or concerns, please commit this change to the
repository when you're ready.

Cheers,
Michael


On Mon, Jul 19, 2010 at 03:18:22PM -0400, Doug Jacobsen wrote:
> Hi Everyone,
> 
> Todd forwarded me a bug fix from some people at LLNL and asked me to take a
> look at it. So far I have gotten through one of the files and I think it is
> actually a bug (at least if I understand the point of the variables
> correctly).
> 
> The file the is modified is src/framework/module_block_decomp.F
> Basically, this is a bug if local_cell_list is supposed to be a list of the
> cells the a specific processor owns. Where in the current implementation on
> a single processor it ends up being an array filled with one processors id
> number. Anyways here is the diff.
> 
> 112,115c112
> <
> <          do i=1,size(local_cell_list)
> <             local_cell_list(i) =  i
> <          end do
> ---
> >          local_cell_list(:) = dminfo % my_proc_id
> 
> 
> Let me know if anyone has any concerns, and if you agree it is indeed a bug
> I can commit a fix.
> 
> Doug

> _______________________________________________
> mpas-developers mailing list
> mpas-developers at mailman.ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/mpas-developers