itaps-parallel Notes from tutorial exercises

[Tim Tautges]

Yep, I'll do that.  Note to others, we "ratified" the move to the 
C-based interface based in part on it working out of the box from 
Fortran.  That does take extra work to achieve due to the underscore 
handling differences on various machines.  I think it will look very bad 
if only one of our interface implementations supports Fortran (like it 
or not, there are MANY scidac apps implemented in that language, maybe 
even more than half of them).

- tim

Devine, Karen D wrote:
> Tim:
> 
> Do you have time to build and run your example programs (FindAdj, List&Sets,
> etc.) that are in the tutorial slides?  In response to Lori's question, I
> tried to build the F77 FindAdjacency example, but I didn't succeed.  I
> couldn't find F77 include files in FMDB.  I tried using MOAB, but cray
> pointers in the example appear to be 32-bit while the MOAB libraries that
> are installed are 64-bit.  I don't know enough about cray pointers and
> fortran to resolve the problem.  Even if I had gotten everything compiled, I
> didn't have an appropriate input file for testing.
> 
> It would be a good sanity test for us to make sure our examples in the
> slides compile and run correctly.  We could have a separate directory for
> them in /usr/local/itaps.
> 
> If I had a working example to use, I think I could convert hello_iMeshP.c to
> hello_iMeshP.F.
> 
> Also, the FindAdj example in the slides has the root_set == 0 assumption in
> at least the C and F77 versions.
> 
> Thanks!
> Karen
> 
> 
> 
> 
> On 11/11/08 11:02 AM, "Lori A. Diachin" <diachin2 at llnl.gov> wrote:
> 
>> Hi Vitus and all,
>>
>> I went through both tutorials this morning and overall think they're
>> quite nice.  I like that you provide the basics here in the second
>> exercise - it's a very nice bridge between Hello and Smooth.  I have a
>> few minor notes (below).
>>
>> As a general comment to everyone - do we have any tutorial exercises
>> that aren't in C++?  I've gone through Hello, Load Balance and Smooth so
>> far and I'm worried we'll give folks the wrong impression..   I see a .c
>> file in the Hello directory -- would it be hard to add a fortran hello
>> example?
>>
>> Lori
>>
>> Tutorial 1:
>>
>> Exercise 2: It's not clear to me where the MPI communicator is
>> initialized in the test file?  I assume it's COMM_WORLD, but you refer
>> to 'initialized above' and I don't see it.  I assume it's in the Init
>> function, but it might be helpful to make it more explicit
>>
>> Exercise 3: Remind users how to get a string length; not all users will
>> be c++ savy.
>>
>> In general it's not always clear to me why I'm doing something, getting
>> the parts, getting various handles, etc - is it needed for Zoltan?  If
>> you can, it would be useful to add a sentence to the beginning of each
>> exercise giving a bit of motivation for why I need to call this
>> function.  It will help folks think through it themselves later..
>>
>>
>> Tutorial 2:
>>
>> Might prefer to have different test file names in tutorial 1 and
>> tutorial 2.... Actually, rather than switching files/directories, you
>> could keep them all part of a single tutorial and just use the #define
>> trick that Karen used in Hello.
>>
>> Is .vtu a vtk format?  I'm not familiar with it and was surprised to see
>> .vtu's... you might want to comment on what to expect in the notes for
>> Exercise 1
>>
>>
>> ~
>>
>>
> 
> 
> 
> 

-- 
================================================================
"You will keep in perfect peace him whose mind is
   steadfast, because he trusts in you."               Isaiah 26:3

              Tim Tautges            Argonne National Laboratory
          (tautges at mcs.anl.gov)      (telecommuting from UW-Madison)
          phone: (608) 263-8485      1500 Engineering Dr.
            fax: (608) 263-4499      Madison, WI 53706