[CPMD-list] problems in starting cpmd

[Axel Kohlmeyer]

On Tue, 9 May 2006, Silviu Zilberman wrote:

SZ> Hi,

hi silviu,

SZ> I downloaded and installed the latest CPMD v3.11.1 on our opteron 
SZ> cluster, using intel v9 compiler (below is the header of the makefile). 
SZ> When running a serial job, I keep getting a message that cpmd.x is not 
SZ> found:
SZ> 
SZ> eyring: [103]> ./cpmd.x amide.in > amide.out&
SZ> [1] 21593
SZ> eyring: [104]> zsh: command not found: ./cpmd.x

<grin>. this is one of the most arcane error messages
on linux machines. people who remember the move from
the a.out to the ELF binary format many, many years
back have seen it a lot, too. this does not mean, that
cpmd.x is not found, but a shared linker or related
compound.

please do: ldd cpmd.x 
to liste the requirements of the binary so we can
see what is missing.

SZ> Then if I run it in parallel I get:
SZ> 
SZ> eyring: [102]> mpirun -np 1 cpmd.x amide.in > amide.out&
SZ> [1] 21533
SZ> eyring: [103]> /opt/mpich.intel/bin//mpirun: 
SZ> /home/silviu/Work/Calculations/CPMD/dipole/cpmd.x: /lib/ld64.so.1: bad 
SZ> ELF interpreter: No such file or directory

that could be the same error.

SZ> Indeed there is no such file /lib/ld64.so.1 but there are other 64 bit 
SZ> ld's in a different directory:
SZ> 
SZ> eyring: [105]> ls /lib64/ld*
SZ> /lib64/ld-2.3.2.so*  /lib64/ld-linux-x86-64.so.2@  
SZ> /lib64/ld-lsb-x86-64.so.1@
SZ> 
SZ> 
SZ> Any ideas how to fix these problems? Is it related to my choice of 
SZ> Blas/Lapack libraries?

have you tried others? mkl?
i'd need to see the output of ldd on the various objects to make sure.

SZ> 
SZ> Thanks, Silviu.
SZ> 
SZ> FFLAGS =  -O -r8

please try, -O2 -unroll
-r8 should not be needed and used anymore for cpmd 3.11.1 we cleaned
up all non-conforming instrinsics during the 3.10 development cycle.

SZ> #LFLAGS = -L/usr/lib64/  -llapack -L/opt/ATLAS/lib/ -lf77blas -lcblas 
SZ> -latlas /opt/acml2.5.0/gnu64/lib/libacml.so -lg2c -static
SZ> LFLAGS = /usr/local/lib/libgoto_opt64-r0.96.so 
SZ> /opt/acml2.5.0/gnu64/lib/libacml.so -lg2c -static

ouch. you should not use '-static' with shared libraries.
please use '-i-static'. this way all compiler related libraries
will be linked statically, but libc/libm and related not
allowing the use of shared libraries otherwise.

also, i suggest to use an adapted libacml.a 

instructions on making the gcc/g77 compile acml compatible
with the intel compilers are in the INSTALL directory 
(if it is not in your cpmd source archive, you;ll find
it somewhere in the contrib section of cpmd.org download)

a short form is at:
https://www.liniac.upenn.edu/wiki/tiki-index.php?page=acml+for+CMM


SZ> CFLAGS =
SZ> CPP = /lib/cpp -P -C -traditional
SZ> CPPFLAGS = -D__alpha -DPOINTER8 -DLAPACK -DFFT_DEFAULT -DALPHALINUX \
SZ>         -DMYRINET -DPARALLEL=parallel -DMP_LIBRARY=__MPI -DLINUX_IA64_INTEL

-DLINUX_IA64_INTEL is wrong for this architecture
(does not resolve to anything anyways).
in case of using MKL (for which this was initially intended)
you have to use now -DINTEL_MKL
to ifdef out the multiply defined IZAMAX.
alternative for that would be to make the linker more permissive
by adding the following to LDFLAGS: -Wl,--allow-multiple-definitions

please let me know, if you need additional help,
   axel.

SZ> NOOPT_FLAG =
SZ> CC = mpicc -O2 -Wall -m64
SZ> FC = mpif90 -c
SZ> LD = mpif90
SZ> AR = ar
SZ> 
SZ> #----------------------------------------------------------------------------
SZ> # Personal Configuration
SZ> #----------------------------------------------------------------------------
SZ> 
SZ> _______________________________________________
SZ> CPMD-list mailing list
SZ> CPMD-list at cpmd.org
SZ> http://cpmd.org/mailman/listinfo/cpmd-list
SZ> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.