[MOAB-dev] MOAB debug

[Grindeanu, Iulian R.]

yes, but the mesh is only 10 mil/ vertices, hexes, 30 mil quads.
so it should "work" up to 2B (with integer )
________________________________________
From: moab-dev-bounces at mcs.anl.gov [moab-dev-bounces at mcs.anl.gov] on behalf of Tim Tautges [timothy.tautges at cd-adapco.com]
Sent: Tuesday, June 23, 2015 2:30 PM
To: moab-dev at mcs.anl.gov
Subject: Re: [MOAB-dev] MOAB debug

Remember global ids are 32-bit...

- tim

On 06/23/2015 02:07 PM, Grindeanu, Iulian R. wrote:
> hmmm
> 4194304 = 2048^2
>
> ________________________________________
> From: Kanaev A.A. [kanaev at ibrae.ac.ru]
> Sent: Tuesday, June 23, 2015 1:57 PM
> To: Grindeanu, Iulian R.
> Subject: FW: MOAB debug
>
> Here's the test program just in case
> ________________________________________
> From: Kanaev A.A.
> Sent: Tuesday, June 23, 2015 9:49 PM
> To: Grindeanu, Iulian R.
> Cc: moab-dev at mcs.anl.gov
> Subject: RE: MOAB debug
>
> Hello Iulian!
> I've reduced the problem with that mesh to some point.
> The mesh now contains no sets and no tags except for GLOBAL_ID and PARALLEL_PARTITION tags
> For every quad i wrote its connectivity nodes global ids as tag NODES with 4 integer variables
> Then i read the mesh on 1 processor but with PARALLEL=READ_PART option, get connectivity for each quad and check if new connectivity nodes' global ids are equal to what i have in NODES tags
> I can see that starting from quad 4194304 connectivity is corrupted as can be seen on the screenshot attached
> Previous tests show that tag data remains uncorrupted but connectivity table somehow shifts if there's PARALLEL=READ_PART option in load_file() function
> Anton
> ________________________________________
> From: Grindeanu, Iulian R. [iulian at mcs.anl.gov]
> Sent: Monday, June 22, 2015 11:40 PM
> To: Kanaev A.A.
> Subject: RE: MOAB debug
>
> Sorry, I didn't look at this problem.
> One issue could be with the sets in the file.
> Do you have many mesh sets?
> Besides the parallel partition sets?
> Can you delete the mesh sets that you do not need; probably you will want material sets and neumann sets, but if you have many geometry sets, it could be a problem.
>
> to enable debug during reading or writing, you need to pass an option, like
>
> -O DEBUG_IO=1 (reading) or -o DEBUG_IO=1 (for writing, with mbconvert)
>
> It does not work very well if you run on too many procs, but on < 30 should be OK
>
> Iulian
> ________________________________________
> From: Kanaev A.A. [kanaev at ibrae.ac.ru]
> Sent: Monday, June 22, 2015 3:26 PM
> To: Grindeanu, Iulian R.
> Subject: MOAB debug
>
> Actually, the issue is a bit more wider
> I have a problem with reading that huge mesh with option PARALLEL even on 1 processor, while it doesn't appear for serial load_file
> Any advices on debugging that?
>
> ________________________________________
> From: Kanaev A.A.
> Sent: Monday, June 22, 2015 11:22 PM
> To: Grindeanu, Iulian R.
> Subject: MOAB debug
>
> Hello Iulian, how can i enable debug output like "dbgOut.tprint(1, "Reading all nodes...\n");"?
> Thanks
> Anton
>

--
Timothy J. Tautges
Manager, Directed Meshing, CD-adapco
Phone: (631) 629-3139 (internal: x29130)
Mobile/Gvoice: (608) 354-1459
timothy.tautges at cd-adapco.com