Re: Re: AMBER: compilation error in SGI Altix 4700 with Intel Fortran10.1.008 and IMKL

From: Ye Mei <>
Date: Fri, 7 Dec 2007 11:23:11 +0800

Hi, Andreas,

Thank you for your reply.
The file config.h generated by configure command is OK. But in the manual of IMKL 10, Intel suggests using mkl_intel_lp64, mkl_intel_thread and mkl_core, instead of the legacy way. I have contacted Intel, and they told me to link the libraries in the following way. And it works for me. It seems to be a bug of IMKL instead of Amber9.
LOADLIB= -lmpi /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ /disk1/chem/intel/mkl/ -liomp5 -ldl

Best regards,
Ye Mei
Institute of Theoretical and Computational Chemistry
Key Laboratory of Mesoscopic Chemistry of MOE
School of Chemistry and Chemical Engineering
Nanjing University
Nanjing 210093

======= 2007-12-06 23:55:12 Andreas Svrcek-Seiler wrote=======

>> I am trying to compile amber9 in our SGI Altix 4700 using Intel Fortran 10.1.008 and IMKL As suggested by Intel, I used libmkl_intel_lp64.a, libmkl_intel_thread.a
>> and libmkl_core.a, instead of the legacy libmkl_lapack.a and
>> libmkl_ipf.a.
>> It failed when linking the libraries. The config file and make.log are
>> I have no idea whether this error is due to any bugs in Amber or IMKL.
>> Can anyone help me?      
>...I try to:
>./configure sgi_altix
>on a linux(!) em64t machine with ifort (and changing the lib-paths in config.h)
>"make" runs through and sander is linked as follows:
><...> ../lib/nxtsec.o ../lib/sys.a
>-L/scr/orgon/svrci/ICC90/intel/mkl/10.0.011/lib/em64t -lguide -lpthread -ldl
> why try other libraries? There's definitely no bug.
>I guess you've either missed at least one library or
>link them in the wrong order (order can matter here).
>good luck

= = = = = = = = = = = = = = = = = = = =


The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun Dec 09 2007 - 06:07:34 PST
Custom Search