[petsc-dev] [petsc-maint #67637] Re: [petsc-users] Scalability of AO ?

[Barry Smith]

  We originally started with multiple libraries because in the 1990's PETSc libraries were "BIG" and made live hard for some filesystems and linkers. Now libpetsc.a is not big, by modern standards of big so is there any reason at all to keep the option of having lots of libraries (that don't get tested properly)?

  Barry

On Mar 21, 2011, at 3:50 PM, Satish Balay wrote:

> --with-single-library=0 doesn't get much testing these days. Perhaps
> its time to get rid of this option..
> 
> satish
> 
> On Mon, 21 Mar 2011, Sebastian Steiger wrote:
> 
>> Oh, I forgot to say I'm using --with-single-library=0 for historical
>> reasons.
>> 
>> Sebastian
>> 
>> 
>> On 03/21/2011 04:44 PM, Satish Balay wrote:
>>> libpetsccharacteristic shouldn't exist even in the real build.
>>> [all objs go into libpetsc.a]. Perhaps something is not in sync?
>>> 
>>> asterix:/home/balay/spetsc/src/characteristic/impls/da>touch slda.c
>>> asterix:/home/balay/spetsc/src/characteristic/impls/da>make
>>> /home/balay/soft/linux64/mpich2-1.1/bin/mpicc -o slda.o -c -fPIC -Wall -Wwrite-strings -Wno-strict-aliasing -Wno-unknown-pragmas -g3   -I/home/balay/spetsc/include -I/home/balay/spetsc/asterix64/include -I/home/balay/soft/linux64/mpich2-1.1/include    -D__INSDIR__=src/characteristic/impls/da/  slda.c
>>> /usr/bin/ar cr /home/balay/spetsc/asterix64/lib/libpetsc.a slda.o
>>> if test -n ""; then /usr/bin/ar cr  slda.lo; fi
>>> /bin/rm -f slda.o slda.lo
>>> asterix:/home/balay/spetsc/src/characteristic/impls/da>
>>> 
>>> Satish
>>> 
>>> 
>>> On Mon, 21 Mar 2011, Sebastian Steiger wrote:
>>> 
>>>> Hong, Jose,
>>>> 
>>>> I have rebuilt petsc-dev and successfully compiled. I also svn update'd
>>>> slepc-dev and compiled the real version. But SLEPc-complex complains
>>>> about not finding libpetsccharacteristic, and indeed it's only there in
>>>> the real version of PETSc. However it's defined in some of the configure
>>>> files and I suspect that's why SLEPc asks for it.
>>>> 
>>>> Am I doing something wrong or is libpetsccharacteristic missing indeed
>>>> in the complex PETSc build?
>>>> 
>>>> Sebastian
>>>> 
>>>> 
>>>> steiger at jaguarpf-login3:~/src/app-nemo/NEMO/libs/petsc-barry/build-cplx$
>>>> grep -inr characteristic conf*
>>>> conf/rules:958:        if [ "${LIBNAME}" = "libpetsccharacteristic" ];
>>>> then OTHERLIBS="${PETSC_DM_LIB_BASIC}"  ; fi;\
>>>> conf/rules:959:        if [ "${LIBNAME}" = "libpetscksp" ]; then
>>>> OTHERLIBS="${HYPRE_LIB} ${PETSC_CHARACTERISTIC_LIB_BASIC}"  ; fi;\
>>>> conf/variables:61:PETSC_CHARACTERISTIC_LIB_BASIC = -lpetsccharacteristic
>>>> ${PETSC_DM_LIB_BASIC}
>>>> conf/variables:62:PETSC_KSP_LIB_BASIC            = -lpetscksp
>>>> ${PETSC_CHARACTERISTIC_LIB_BASIC}
>>>> conf/variables:72:PETSC_CHARACTERISTIC_LIB = ${C_SH_LIB_PATH}
>>>> -L${PETSC_LIB_DIR} ${PETSC_CHARACTERISTIC_LIB_BASIC}
>>>> ${PETSC_EXTERNAL_LIB_BASIC}
>>>> conf/variables:92:SHLIBS   = libpetscsys libpetscvec libpetscmat
>>>> libpetscdm libpetsccharacteristic libpetscksp libpetscsnes libpetscts
>>>> libpetsccontrib
>>>> config/PETSc/Configure.py:256:
>>>> self.addMakeMacro('PETSC_CHARACTERISTIC_LIB','${C_SH_LIB_PATH}
>>>> ${PETSC_WITH_EXTERNAL_LIB}')
>>>> config/PETSc/Configure.py:267:
>>>> self.addMakeMacro('PETSC_CHARACTERISTIC_LIB','${PETSC_WITH_EXTERNAL_LIB}')
>>>> config/PETSc/Configure.py:294:    fd.write('\"Machine characteristics:
>>>> %s\\n\"\n' % (platform.platform()))
>>>> Binary file config/PETSc/Configure.pyc matches
>>>> config/builder.py:858:       dirs = map(lambda d: os.path.join(rootDir,
>>>> 'src', d), ['inline', 'sys', 'vec', 'mat', 'dm', 'ksp', 'snes', 'ts',
>>>> 'characteristic', 'docs', 'tops'])
>>>> config/cmakegen.py:152:                 ('characteristic' , 'dm vec sys'),
>>>> config/manifest.in:12:recursive-exclude src/characteristic/examples *
>>>> config/BuildSystem/config/setsBackport.py:10:characteristic operation
>>>> for sequences is subscripting with small
>>>> config/BuildSystem/config/setsOrdered.py:11:characteristic operation for
>>>> sequences is subscripting with small
>>>> Binary file config/BuildSystem/config/setsOrdered.pyc matches
>>>> Binary file config/cmakegen.pyc matches
>>>> 
>>>> 
>>>> steiger at jaguarpf-login3:~/src/app-nemo/NEMO/libs/petsc-barry/build-cplx$
>>>> ls characteristic linux-complex/lib/*
>>>> ls: characteristic: No such file or directory
>>>> linux-complex/lib/libblacs.a         linux-complex/lib/libparmetis.a
>>>> linux-complex/lib/libpetscmat.a    linux-complex/lib/libpetscsys.so
>>>> linux-complex/lib/libpord.a
>>>> linux-complex/lib/libcmumps.a        linux-complex/lib/libpetscdm.a
>>>> linux-complex/lib/libpetscmat.so   linux-complex/lib/libpetscts.a
>>>> linux-complex/lib/libscalapack.a
>>>> linux-complex/lib/libdmumps.a        linux-complex/lib/libpetscdm.so
>>>> linux-complex/lib/libpetscsnes.a   linux-complex/lib/libpetscts.so
>>>> linux-complex/lib/libsmumps.a
>>>> linux-complex/lib/libmetis.a         linux-complex/lib/libpetscksp.a
>>>> linux-complex/lib/libpetscsnes.so  linux-complex/lib/libpetscvec.a
>>>> linux-complex/lib/libsuperlu_dist_2.5.a
>>>> linux-complex/lib/libmumps_common.a  linux-complex/lib/libpetscksp.so
>>>> linux-complex/lib/libpetscsys.a    linux-complex/lib/libpetscvec.so
>>>> linux-complex/lib/libzmumps.a
>>>> 
>>>> 
>>>> 
>>>> On 03/21/2011 11:38 AM, Jose E. Roman wrote:
>>>>> 
>>>>> El 21/03/2011, a las 16:16, Barry Smith escribió:
>>>>> 
>>>>>> On Mar 21, 2011, at 10:01 AM, Sebastian Steiger wrote:
>>>>>> 
>>>>>>> Hong,
>>>>>>> 
>>>>>>> Thanks a lot for the effort. Unfortunately I have a problem testing the
>>>>>>> stuff (that also applies to issue with the ASM preconditioner that Barry
>>>>>>> is working on): I'm using a rather complicated build system that
>>>>>>> incorporates real+complex PETSc+SLEPc. It seems to me that the current
>>>>>>> SLEPc release does not work anymore with petsc-dev.
>>>>>>> 
>>>>>>> So the only thing I could do is somehow compile a separate program that
>>>>>>> reads matrices and vectors from file that are generated by my simulation
>>>>>>> and then use that with petsc-dev. But that requires a few days if not
>>>>>>> weeks of work for me, and I hate to say that my tight schedule is tight
>>>>>>> until May.
>>>>>>> 
>>>>>>> So I will probably have to wait with the testing on my side until
>>>>>>> petsc-3.2 is out and then maybe another month until Jose and the SLEPc
>>>>>>> guys have an update of SLEPc. Do you see another option? I'd be glad to
>>>>>>> do some testing if it requires not too much effort.
>>>>>>> 
>>>>>>   It's best to just try to get SLEPc-dev working with PETSc-dev and do your testing in the proper place.
>>>>>> 
>>>>>>   Maybe Jose can tell us what needs updateing in SLEPc and someone can fix it if Jose doesn't have time at the moment.
>>>>>> 
>>>>>>  Barry
>>>>>> 
>>>>> 
>>>>> I have just committed a change to synchronize slepc-dev with petsc-dev. Let me know if problems arise.
>>>>> 
>>>>> Barry, there may be an inconsistency in the names of PETSC_REAL_MAX and PETSC_MIN_REAL.
>>>>> 
>>>>> Jose
>>>>> 
>>>> 
>>>> 
>>>> 
>> 
>> 
>>