itaps-parallel /usr/local/itaps on mesh.ep.wisc.edu

[Mark Miller]

Hi Tim,

Ok, tried that. If fails in the same way. So, I think the problem is
related to 64 bit compatibility in FMDB's ITAPS implementation. If I am
reading the error messages and header files right, at some point, it is
trying to cast a void* to an unsigned int...

iMesh/iMesh_FMDB.cc: In function ‘void imesh_destroytag_(void*, void*, int, int*)’:
iMesh/iMesh_FMDB.cc:3575: error: cast from ‘void*’ to ‘pMeshDataId’ loses precision

which would work fine on 32 bit but not on 64 bit.

I suppose I could try compiling everything for 32 bit but that would
mean everything including all VisIt and its libs like VTK, Qt, Mesa,
Silo, etc. and so I am recluctant to spend time on that now.

Mark


On Thu, 2008-09-11 at 12:44 -0500, Tim Tautges wrote:
> Mark, you can use 4.2 series by explicitly calling for gcc-4.2, and I 
> think code compiled with that series will link with other code compiled 
> with 4.3.
> 
> - tim
> 
> Mark Miller wrote:
> > Hi All,
> > 
> >    Ok, I've installed plugins for MOAB and GRUMMP implementations to
> > the currently installed VisIt in /usr/local/itaps and confirmed they
> > work on test data. The GRUMMP plugin works only on volume mesh files. I
> > don't know why yet but it does not work on surface mesh (.smesh) or
> > boundary (.bdry) files for GRUMMP. These plugins are named 'ITAPS_MOAB'
> > and 'ITAPS_GRUMMP'.
> > 
> >    I tried building the FMDB installation I have installed locally at
> > LLNL, fmdb_rev2272.tar.gz, on mesh.ep.wisc.edu. I ran into problems that
> > I think are related to gcc-4.3 compatibility so I did not persue
> > further. So, there is presently no FMDB plugin for VisIt.
> > 
> > Mark
> > 
> > On Thu, 2008-09-11 at 10:30 -0500, Tim Tautges wrote: 
> >> Mark Miller wrote:
> >>> Hi All,
> >>>
> >>> I've been working on getting VisIt RE-installed with support for all
> >>> three ITAPS implementations and have a hit a couple of snags...
> >>> 2. -fPIC not used to produce the .a's for MOAB. I need to build
> >>>    .so's for VisIt. To do this from .a's for MOAB, those .a's needed
> >>>    to have been compiled with position independent code (-fPIC).
> >>>    Alternatively, if .so's for MOAB are there, I can use those
> >>>    directly.
> >>>
> >> Ok, I've modified the installed MOABser on mesh.ep to have both shared 
> >> and static libs.  For parallel does it make sense to have shared libs? 
> >> MOAB's make didn't work for parallel with shared libs, due to hdf5 
> >> having unresolved externals.
> >>
> >> - tim
> >>
> 
-- 
Mark C. Miller, Lawrence Livermore National Laboratory
email: mailto:miller86 at llnl.gov
(M/T/W) (925)-423-5901 (!!LLNL BUSINESS ONLY!!)
(Th/F)  (530)-753-8511 (!!LLNL BUSINESS ONLY!!)