itaps-parallel testing the Smooth tutorial

[Devine, Karen D]

Martin:  Nice example!  I had no problems understanding, building or running
the code.  My biggest questions are about the results.  See below.

Some comments:

-  In the results, do you expect the serial smooth_iMesh.cc to give
different results from smooth_iMeshP.cc run on one processor?  If so, you
should probably comment on that in the instructions; if not, I did something
wrong.
iMesh:   Inverse Mean Ratio     1.01211     1.27264     1.28057     1.81256
0.142304
iMeshP:  Inverse Mean Ratio     1.01226     1.27227     1.28018     1.81257
0.142116


-  In the results, do you expect the metrics for one pass to differ when run
on one processor compared to runs on four processors?   If so, you should
probably comment on that in the instructions; if not, I did something wrong.
ONE PROC:    Inverse Mean Ratio     1.01226     1.27227     1.28018
1.81257    0.142116
FOUR PROCS:  Inverse Mean Ratio     1.01436     1.27575     1.28417
1.8822     0.14682


-  I tried the multiple-pass runs that you suggested.  I don't understand
the output there.  For the two-pass run, I get these quality metrics after
each pass:
Inverse Mean Ratio     1.01358     1.27586     1.28423      1.8822
0.146376
Inverse Mean Ratio     1.00785     1.26025     1.26761       1.785
0.13641

For the three-pass run, I get these quality metrics after each pass:
Inverse Mean Ratio     1.01384     1.27553     1.28392      1.8822
0.14654
Inverse Mean Ratio     1.00571     1.25996      1.2674     1.78658
0.137086
Inverse Mean Ratio     1.00678     1.25397     1.26125     1.73134
0.135249

My question:  Why does the output of the first and second passes differ
between the two-pass run and the three-pass run?  I expected the first two
passes of the three-pass run to match the output of the two-pass run.


-  In the instructions, you explain the output from the mesh smoothing
application.  (That's good!)  You include the following phrase:  "... while
the quality of the worst-shaped element went from 2.45977 down to 1.81256."
Usually, decreasing quality is a bad thing, not a desired output.  Is it the
quality of the element going down, or some other metric (that we'd like to
decrease)?


-  I corrected a few typos in the instructions in my copy; you can diff and
copy them if you like:  ~kddevin/Smooth/HandsOn_Session2_Smooth.txt .  They
are quite minor, however, so feel free to ignore them.

Karen





On 11/7/08 3:54 PM, "Martin Isenburg" <isenburg1 at llnl.gov> wrote:

> hello,
>
> does anyone have a few minutes to spare to try out the "Smooth"
> tutorial on /usr/local/itaps/Smooth?
>
> let me know if you have any comments ...
>
> martin
>
>
>