[Octopus-users] Compilation with AIX 5 compiler (fwd)
Francesco Sottile
Francesco.Sottile at polytechnique.edu
Thu Apr 16 10:16:14 WEST 2009
> Dear Octopus developpers,
>
> I am trying to compile Octopus-3.1-0 with AIX 5L compiler
> (IBM/Hitachi), but I encounter several errors during
> the ./configure step – conf.log is attached. The same errors
> occur with previous Octopus-3.0. Specifically,
> the errors are:
>
> 1. The first error I encounter is about a.out, the log
> messages show:
> checking for C compiler default output file name... a.out
> checking whether the C compiler works... configure: error: cannot
> run C compiled programs. If you meant to cross compile, use `--host'.
>
> 2. By adding --host=powerpc in the ./configure file, I can skip
> the above item,
I'm afraid this doesn't solve anything and you should already care about
this first problem, for it seems you're not able to create a working
executable. Putting --host just tells the configure procedure not to run
the executables (because they're meant to run on another architecture) but
the executables will be wrong anyway.
try to pass the configure a c compiler
./configure CC=gcc
or
./configure CC=xlc
and check if the error is still there
> and arrive at the following error:
> checking for numa_available in -lnuma... no
> checking if malloc aligns memory to 16 byte boundaries... configure:
> error: cannot run test program while cross compiling
>
this error is linked to the first one. you shouldn't use --host
> 3. By commenting out lines of malloc checking, then I proceed
> and get:
> checking how to get verbose linking output from mpif90... -###
> checking for Fortran libraries of mpif90... -L/usr/lpp/ppe.poe/lib
> -L/usr/lpp/ppe.poe/lib/ip -lmpi_
> r -L/opt/hitachi/matmpp/lib -L/opt/hitachi/matsss/lib
> -L/opt/hitachi/MSL2/lib -L/usr/local/lib -l/tm
> p/.20090416082000360500F90/conftest/conftest.L -L/opt/ofort90/lib
> -lhf90vecmath -lhf90math -lf90_r -lsmp -lfiexpmpi_r -lpthreads
> -lc_r
> checking for dummy main to link with Fortran libraries... unknown
> configure: error: linking to Fortran libraries from C fails
>
and also this. I suggest to solve the above problems first f.
> 4. By commenting out the above library checking lines in
> the ./configure file, I can proceed to the following point.
> However, I have checked that the sgemm package does exist
> in the BLAS routine.
>
> checking if the Fortran compiler uses malloc... no
> **********************************************
> *** octopus_mpi will be generated
> **********************************************
> some parameters were not parsed.
> checking for sgemm... no
> checking for ATL_xerbla in -latlas... no
> checking for sgemm in -lmkl_ia32... no
> checking for sgemm in -lmkl_def... no
> checking for sgemm in -lblas... no
> checking for acosp in -lsunmath... no
> checking for sgemm in -lblas... no
> checking for sgemm in -lgoto... no
> checking for sgemm in -lmkl... no
> checking for sgemm in -lcxml... no
> checking for sgemm in -ldxml... no
> checking for sgemm in -lscs... no
> checking for sgemm in -lcomplib.sgimath... no
> checking for sgemm in -lblas... no
> checking for sgemm in -lblas... no
> configure: error: could not find required blas library
>
> I quit further attempts here, and I would like to have
> your advices about how to go around these errors.
> Thank you.
>
> Best regards,
> M.Tanaka
>
>
,ø€º°`°º€ø,žž,ø€º°`°º€ø,žž,ø€º°`°º€ø,žž,ø€º°`°º€ø,žž,ø€º€ø,
) (
( Francesco Sottile )
) email: francesco.sottile at polytechnique.edu (
( web: http://theory.polytechnique.fr/people/sottile/sito/ )
) (
( Laboratoire des Solides Irradies (LSI) )
) Ecole Polytechnique 91128 Palaiseau cedex, France (
( Tel. 0033-1-69334549 (Fax. 0033-1-69334543) )
) (
( European Theoretical Spectroscopy Facility (ETSF) )
) http://www.etsf.eu (
( )
€ø,žž,ø€º°`°º€ø,žž,ø€º°`°º€ø,žž,ø€º°`°º€ø,žž,ø€º€ø,žž,ø€º°`
More information about the Octopus-users
mailing list