[MOAB-dev] Parallel mesh read results

[Dmitry Karpeev]

Forgot to mention this: in the spreadsheet file the cells that
indicate failure are colored
according the the (conjecture) type of failure. In addition, each
"failure" cell has a note indicating the
error MOAB generates in that case (all errors were the same on all
procs, so only one
is quoted).

Dmitry.

On Mon, Apr 12, 2010 at 2:59 PM, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:
> Here are some results on my parallel mesh read tests.
> In this spreadsheet I only indicate success/failure; the actual
> running times will come a bit later.
> I have verified the entries on this spreadsheet on my laptop and on
> cosmea running on 4 processes each time.
> These results include the following meshes: blocks-64.h5m,
> tjunc6_64.h5m, tjunc6RIB_16.h5m.
>
> I'm now running results for a large mesh (br64_1m_1024.h5m -- 64bricks
> with 1m hexes, partitioned 1024 ways).
> The reason this mesh is not included here is that it exhibits a
> different pattern of success/failure, which I'm trying
> to identify and ensure it is independent of the hardware I run on
> (laptop vs cosmea).
>
> Beyond this, the next step is to look at the actual source code
> closely to try to identify the root cause of the failures.
>
> Thanks.
> Dmitry.
>