[Pw_forum] RE : Re: problem compiling espresso with cmkl libraries

[Axel Kohlmeyer]

On Fri, 5 Oct 2007, halim said wrote:

HS> Dear Axel,
HS>    
HS>   Thanks lot for your quick response, I have edited the make.sys and added  '-lsvml' after '-lguide'
HS>    

HS>   and make clean before recompiling again, the error message is the
HS> same as without adding
HS>    '-lsvml' after '-lguide'.  The compiler that I use is intel
HS> fortran 9.0, is there other alternative way to success compiling the
HS> code.

i am not convinced. but here is one alternative. please change FFLAGS to:

FFLAGS         = -O2 -tpp6 -assume byterecl

...and as paolo indicated, please also make sure that you have the
latest patchlevel of the intel compiler. ever since version 7.x all
initial releases of the intel compilers had some serious deficiencies,
that were only resolved after multiple patches. the finaly 9.0 
patchlevel was quite ok, but both, 8.1 and 9.1 are more reliable.


HS>   Your help and advice are very appreciate.
HS>    

HS>   PS: I have also linked the code by blas and lapack of
HS> esspresso.3.2 code but the errors messages appear also for linking.

those errors have nothing to do with blas/lapack.

cheers,
   axel.


HS>    
HS>   Yours sincerely,
HS>    
HS>   Halim
HS> 
HS> Axel Kohlmeyer <akohlmey at cmm.chem.upenn.edu> a écrit :
HS>   On Thu, 4 Oct 2007, halim said wrote:
HS> 
HS> HS> 
HS> HS> Dear All,
HS> 
HS> dear halim,
HS> 
HS> HS> I am trying to compile espresso3.2 in serial version, with ifort
HS> HS> compiler and cmkl library in cluster machine.
HS> HS> When compiling pw alone as test I got the errors messages, could
HS> HS> you please help me in order to solve my problem.
HS> 
HS> first off, thanks for providing a complete problem report
HS> with all relevant information.
HS> 
HS> HS> First:
HS> HS> 
HS> HS> ./configure F90=ifort
HS> HS> 
HS> HS> Second
HS> HS> 
HS> HS> make pw
HS> HS> 
HS> HS> these the errors messages that I got, I do not know if the origin is from linking with cmkl library or related to some think else.
HS> HS> 
HS> HS> -----------------------------------------
HS> HS> ../Modules/ptoolkit.o ../Modules/random_numbers.o ../Modules/read_cards.o 
HS> HS> ../Modules/read_namelists.o ../Modules/read_upf.o ../Modules/read_uspp.o 
HS> HS> ../Modules/recvec.o ../Modules/shmem_include.o ../Modules/splinelib.o 
HS> HS> ../Modules/stick_base.o ../Modules/task_groups.o ../Modules/timestep.o 
HS> HS> ../Modules/uspp.o ../Modules/version.o ../Modules/wavefunctions.o 
HS> HS> ../Modules/wave_base.o ../Modules/xml_io_base.o libpw.a ../flib/ptools.a ../flib/flib.a ../clib/clib.a ../iotk/src/libiotk.a -lmkl_lapack -L/opt/intel/cmkl/8.0/lib/em64t -lmkl_em64t -lguide -lpthread 
HS> HS> 
HS> HS> IPO Error: unresolved : __svml_cos2
HS> HS> Referenced in clib.a(fft_stick.o)
HS> HS> IPO Error: unresolved : __svml_sin2
HS> HS> Referenced in clib.a(fft_stick.o)
HS> HS> .../clib/clib.a(fft_stick.o)(.text+0x37044): In function `complete_twiddle':
HS> HS> : undefined reference to `__svml_cos2'
HS> HS> .../clib/clib.a(fft_stick.o)(.text+0x37062): In function `complete_twiddle':
HS> HS> : undefined reference to `__svml_sin2'
HS> 
HS> 
HS> this means, that your compiler optimized to use the SSE 
HS> version of sine, cosine etc. those are in the SVML library.
HS> 
HS> just edit your make.sys file to append '-lsvml' after '-lguide'
HS> 
HS> cheers,
HS> axel.
HS> 
HS> HS> -------------------------------------------------------------------
HS> HS> 
HS> HS> Your advices and help are appreciated.
HS> HS> 
HS> HS> Yours sincerely,
HS> HS> 
HS> HS> Halim Said,
HS> HS> dept of Phys, Univ Soussa,
HS> HS> Tunisi
HS> HS> 
HS> HS> 
HS> HS> ---------------------------------
HS> HS> Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail 
HS> 
HS> 

-- 
=======================================================================
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.