Victor,
If you are using mpif90 for compilation and linking, you shouldn't need to
link the mpich library explicitly. Did you also verify that your hdf5 was
compiled using the right versions of compilers and mpi includes ?
On Thu, 12 Aug 2004, Victor Shum wrote:
> Hi Tomak and flash users,
>
> I removed the -fast, -ipo, -ipo_obj and the no action
> performed warning disappeared!
> Thanks a lot!
>
> However, on the link side I still have problems after
> doing clean reinstall of mpich and hdf5.
>
> I did notice that I need the syntax:
> LIB_HDF5 = -L/usr/src/hdf5-1.4.3/src/.libs -lhdf5
>
> for mpif90 to link the hdf5 library without the
> undefined reference hdf5 variable errors but if I try
> to mimic that and use :
> LIB_HDF5 = -L/usr/local/mpich-1.2.5/lib -lmpich
> it didn't work!
> What is the syntax to -lmpich (I just guessed this one
> as -lhdf5 is to the libhdf5.a, I have in the mpich dir
>
> the library libmpich.a) .
>
> I still have those mpi_wtime undefined errors.
>
> Any help on this is most welcome!
>
> Much Appreciated!
>
> Bst,
> Vic
> --- Tomasz Plewa <tomek@flash.uchicago.edu> wrote:
>
> > Victor -
> >
> > I think it might be a good idea to cleanly install
> > intel compiler,
> > recompile mpich, make sure the output from
> >
> > mpif90 -show
> >
> > points to the right place, that your /etc/ld.so.conf
> > points to
> > where you installed your libraries (if you use
> > shared libraries),
> > eventually set up and compile the code.
> >
> > Did you install mpich in
> >
> > /usr/src/mpich-1.2.5.2/include
> >
> > this is rather unusual place reserved for
> > distribution sources rather
> > than compiled tools. Try
> >
> > ./configure --prefix=/usr/local/mpich-1.2.5.2-ifc80
> >
> > Make sure you defined environemnt prior to
> > installation as follows:
> >
> > setenv FC ifort
> > setenv F77 ifort
> > setenv F90 ifort
> > setenv F9X ifort
> > setenv CC icc
> > setenv CXX icc
> >
> > and you may need
> >
> > setenv RSHCOMMAND ssh
> >
> > depending whether rsh or ssh is available on your
> > system.
> >
> > You may have followed all those steps already, but I
> > am finding
> > your diagnostic information a little difficult to
> > decipher.
> >
> > Also, can you lower optimization options and at
> > least temporarily
> > eliminate -fast -ipo -ipo_obj.
> >
> > Best -
> >
> > Tomek
> > --
> > On Thu, Aug 12, 2004 at 04:57:09PM -0700, Victor
> > Shum wrote:
> > >
> > > Hi,
> > >
> > > I checked the include library for
> > > mpich-1.2.5.2/include
> > > and I can see the definitions of MPI_WTIME etc.
> > there
> > >
> > > In my compile script I used:
> > > mpif90 -c -fast -ipo -ipo_obj -r8 -i4 -I
> > > /usr/src/mpich-1.2.5.2/include -DN_DIM=2
> > > -DMAXBLOCKS=1000 -DNXB=8 -DNYB=8 -DNZB=1 xx.f90
> > >
> > > so the mpif.h should have been incorporated right?
> > >
> > >
> > > Also I still get the no action performed warning
> > for a
> > > lot of files, maybe that is related? :
> > >
> > > mpif90 -c -fast -ipo -ipo_obj -r8 -i4 -I
> > > /usr/src/mpich-1.2.5.2/include -DN_DIM=2
> > > -DMAXBLOCKS=1000 -DNXB=8 -DNYB=8 -DNZB=1
> > > get_argument.F90
> > > ifort: Command line warning: no action performed
> > for
> > > specified file(s)
> > > mpif90 -c -fast -ipo -ipo_obj -r8 -i4 -I
> > > /usr/src/mpich-1.2.5.2/include -DN_DIM=2
> > > -DMAXBLOCKS=1000 -DNXB=8 -DNYB=8 -DNZB=1
> > > current_date_time.F90
> > > ifort: Command line warning: no action performed
> > for
> > > specified file(s)
> > > mpif90 -c -fast -ipo -ipo_obj -r8 -i4 -I
> > > /usr/src/mpich-1.2.5.2/include -DN_DIM=2
> > > -DMAXBLOCKS=1000 -DNXB=8 -DNYB=8 -DNZB=1
> > get_opt.F90
> > >
> > >
> > > --- Tomasz Plewa <tomek@flash.uchicago.edu> wrote:
> > >
> > > > Victor -
> > > >
> > > > It is likely that you are linking against wrong
> > > > library (static
> > > > vs. shared, for example).
> > > >
> > > > Tomek
> > > > --
> > > > On Thu, Aug 12, 2004 at 04:07:14PM -0700, Victor
> > > > Shum wrote:
> > > > >
> > > > >
> > > > > Thanks for the info.
> > > > >
> > > > > I did that and still it was the same.
> > > > >
> > > > > When linking, the make did :
> > > > > mpif90 -o flash2 Cosmology.o
> > > > CosmologicalFunctions.o
> > > > > ..
> > > > > flashModules.o -L
> > /usr/src/hdf5-1.4.3/src/.libs
> > > > > -lhdf5
> > > > >
> > > > > However, after linking I got many undefined
> > > > reference
> > > > > :
> > > > >
> > > > > amr_morton.o(.text+0x3356): undefined
> > reference to
> > > > > `mpi_irecv_'
> > > > > amr_morton.o(.text+0x33ae): undefined
> > reference to
> > > > > `mpi_ssend_'
> > > > > amr_morton.o(.text+0x33fb): undefined
> > reference to
> > > > > `mpi_waitall_'
> > > > > amr_morton.o(.text+0x34c3): undefined
> > reference to
> > > > > `mpi_barrier_'
> > > > > amr_prolong_cc.o: In function
> > `amr_prolong_cc_':
> > > > > amr_prolong_cc.o(.text+0x49): undefined
> > reference
> > > > to
> > > > > `mpi_type_vector_'
> > > > > amr_prolong_cc.o(.text+0x6f): undefined
> > reference
> > > > to
> > > > > `mpi_type_commit_'
> > > > > amr_prolong_cc.o(.text+0x174): undefined
> > reference
> > > > to
> > > > > `mpi_irecv_'
> > > > > amr_prolong_cc.o(.text+0x225): undefined
> > reference
> > > > to
> > > > > `mpi_ssend_'
> > > > >
> > > > > I'm using mpich1.2.5.2 and inside the include
> > dir
> > > > > I don't see mpi_ssend when I did a grep
> > mpi_ssend
> > > > *
> > > > >
> > > > > Am I using a wrong version with FLASH2.3?
> > > > >
> > > > > Do I need to re-download a higher version of
> > FLASH
> > > > for
> > > > > the variables to be in the .h files?
> > > > >
> > > > > Many Thanks for your attention!
> > > > >
> > > > > Bst,
> > > > > Victor
> > > > >
> > > > > --- tangsk@astro.umass.edu wrote:
> > > > >
> > > > > > Hi,
> > > > > > I once meet such problem. The solution
> > is
> > > > add
> > > > > > -ipo -ipo_obj after the -fast
> > > > > > optimization. You can refer to the manual of
> > > > intel
> > > > > > fortan compiler for detail.
> > > > > >
> > > > > > Good luck
> > > > > >
> > > > > > Best wishes,
> > > > > > Shikui
> > > > > >
> > > > > > Quoting Victor Shum <vshum12@pacbell.net>:
> > > > > >
> > > > > > > Hi,
> > > > > > >
> > > > > > > Also, I was announcing the successful
> > > > compilation
> > > > > > too
> > > > > > > early since I actually got a lot of these
> > no
> > > > > > action
> >
> === message truncated ===
>
-- Anshu Dubey Astronomy & Astrophysics University of Chicago 5640 S. Ellis Ave. Chicago IL 60637 Tel : (773) 834 2999 Fax : (773) 834 3230Received on Thu Aug 12 22:10:54 2004
This archive was generated by hypermail 2.1.8 : Thu Aug 31 2006 - 21:20:48 CDT