[PETSC #19456] PETSc with blocksolve95 and matlab interface

[Barry Smith]

    I concur, rip it out of petsc-dev

     Barry

    Hmmm, maybe we should rip it out of petsc-3.0.0 but since it  
involves lots of changes, hmmmm

On May 19, 2009, at 1:59 PM, Matthew Knepley wrote:

> I vote for kicking out BlockSolve95. It has not been maintained  
> since 95. It does parallel
> ICC(0) and ILU(0). The ILU can be done by PILUT. It appears the only  
> thing we give up
> is ICC(0) in parallel. IS this worth keeping this headache?
>
>   Matt
>
> On Tue, May 19, 2009 at 1:25 PM, Satish Balay <petsc- 
> maint at mcs.anl.gov> wrote:
> On Tue, 19 May 2009, Rebecca Hartman-Baker wrote:
>
> > Hi,
> >
> > I'm installing PETSc 3.0.0 p4 on jaguar at ORNL, which runs linux  
> on the
> > front end and CNL 2.1 on the compute nodes.  Our MPI is the Cray
> > distribution.  I am compiling with PGI compiler 8.0.3.  The user  
> requested
> > that I create an installation that includes hdf5, hypre, sundials,
> > blocksolve95, and the matlab interface.  In creating this custom
> > installation, I'm running into two problems.
> >
> > First, I was having some trouble figuring out how to get  
> blocksolve95.  When
> > I click on the links in the PETSc website referencing  
> blocksolve95, I get a
> > "page not found" error.  After asking a colleague about it, I was  
> able to
> > track down the ftp site to retrieve it (ftp.mcs.anl.gov/pub/BlockSolve95/) 
> ,
> > but this is one of the few packages where there's an interface to  
> it but no
> > means for PETSc to download and install it for you.  I wanted to  
> make sure
> > that since there doesn't appear to be an official source that I've  
> indeed
> > fetched the right thing.
>
> Yes - we don't have a --download-blocksolve95 option. It has to be
> installed manually - and then specified to PETSc
> configure. [BlockSolve uses very-old install style of petsc - and was
> never updated..]
>
> > Second, when I compile the matlab interface, it appears that  
> there's an
> > invalid flag, -pthread, which is not understood by the PGI  
> compiler.  It
> > appears to be a standard GNU flag that allows the use of  
> pthreads.  I wasn't
> > able to track down where this flag is set.  Other than this, the  
> compilation
> > (without blocksolve95) went just fine.  The logfiles told me to  
> send them to
> > you, so attached please find make.log and configure.log from my  
> compilation.
>
> >>>>>>>
> libfast in: /sw/xt5/petsc/3.0.0/cnl2.1_pgi8.0.3_custom/petsc-3.0.0- 
> p4/src/sys/viewer/impls/socket/matlab
> BEGINNING TO COMPILE MATLAB INTERFACE
> /opt/cray/xt-asyncpe/2.3/bin/cc: INFO: linux target is being used
> /opt/cray/xt-asyncpe/2.3/bin/cc: INFO: linux target is being used
> /opt/cray/xt-asyncpe/2.3/bin/cc: INFO: linux target is being used
> pgcc-Error-Unknown switch: -pthread
>
>    mex: link of ' "sread.mexa64"' failed.
>
> make[10]: [sread] Error 1 (ignored)
> /bin/mv: cannot stat `sread.mex*': No such file or directory
> make[10]: [sread] Error 1 (ignored)
> <<<<<<<<
>
> The errors are from /sw/xt5/matlab/7.7/sles_binary/bin/mex - as we are
> attempting to invoke it with "mex CC=cc" - but perhaps it only works
> with "mex CC=gcc"?
>
>
> Perhaps you can edit src/sys/viewer/impls/socket/matlab/makefile - and
> remove all occurences of "CC=${CC}" - and see if it works.
>
> Satish
>
> >
> > Thanks for your help!
> >
> > Sincerely,
> > Rebecca Hartman-Baker
> >
> >
>
>
>
>
> -- 
> What most experimenters take for granted before they begin their  
> experiments is infinitely more interesting than any results to which  
> their experiments lead.
> -- Norbert Wiener