AMBER Archive (2008)

Subject: Re: AMBER: Amber9 Installation Problem

From: trudy_at_uoguelph.ca
Date: Thu Mar 20 2008 - 17:43:48 CDT

Thanks very much for the suggestions. I had already done the patching
before compiling. As it turns out the system was not configured to use
the LD_LIBRARY_PATH by default (new system) and I hadn't double
checked that before compiling. So, that's fixed now. However, ifort
version 10 still does not work for the same reasons, but the 9.1.x's
now work. The tests run successfully now without the several major
failures that I mentioned previously. The ptraj test runs without
any problems.

I have one more question though.

Previously, I had gotten errors of this magnitude with g95:

TEST_FAILURES.diff file:

possible FAILURE: check cytosine.out.dif
/usr/local/Amber/amber9/test/cytosine
99,100c99,100
< NSTEP = 2 TIME(PS) = 5420.004 TEMP(K) = 300.46
PRESS = -32.0
< Etot = -608.0823 EKtot = 1060.4055 EPtot =
-1668.4878 

---
>  NSTEP =        2   TIME(PS) =    5420.004  TEMP(K) =   300.49   
> PRESS =    33.2
>  Etot   =      -664.1291  EKtot   =      1060.4894  EPtot      =      
> -1724.6185

Notice the large differences between the Etot values.

With the current apparently successful installation using ifort, most  
of the errors are the insignificant ones at the fourth decimal place.  
For example EGB:

possible FAILURE:  check mdout.jar.dif
/usr/local/Amber/amber9/test/jar
180c180
<  EELEC  =        79.1435  EGB     =     -2602.0110  RESTRAINT  =      
     0.0815
---
>  EELEC  =        79.1435  EGB     =     -2602.0109  RESTRAINT  =      
>     0.0815
---------------------------------------

The only exception is this one:

possible FAILURE:  check amoeba_jac.mdout.dif
/usr/local/Amber/amber9/test/amoeba_jac
206c206
<  EKCMT  =         0.  VIRIAL  =         0.  VOLUME     =         0.0013
---
>  EKCMT  =         0.  VIRIAL  =         0.  VOLUME     =         0.0028

Do you think that the differences in volume here is significant?

Quoting Cenk Andac <cenk_andac_at_yahoo.com>:

> One more though,
>   please make sure that your gcc is x86_64 but not x86.
>   gcc -v      should tell you about the version.
>
>   regards,
>   jenk
>
> Cenk Andac <cenk_andac_at_yahoo.com> wrote:
>     well, I have just noticed that the version of gcc you use is not  
>  compatible with the ifort version you are using.
>   Sometimes ifort v9.x.x seems to be more compatible with older   
> versions of gcc.
>   and i am not so sure if you can still download ifort v9 from   
> intel's website.
>
>   regards,
>
>   jenk
>
>
> Cenk Andac <cenk_andac_at_yahoo.com> wrote:
>     Hi,
>
>   I usually patch bugfixes before installing AMBER.
>   I have not tried installing AMBER using g95 before. Perhaps   
> someone else may help you out with g95 installation of AMBER.
>   As for the installation of AMBER using ifort and mkl :
>   1) make sure that LD_LIBRARY_PATH is set-up correctly for MKL and   
> IFORT libraries.
>   2) MKL_HOME is set correctly
>   3) A path is set to ifort's binary directory and also make sure   
> that  ifort is sourced
>   4) if your cpu's are intel- or AMD64-based try the following configuration
>      ./configure -bintraj -p4 ifort_x86_64
>      you may also try a static configuration as you have tried before.
>
>
>   good luck
>
>   jenk
>
> trudy_at_uoguelph.ca wrote:
>   Hello all,
>
> I am currently trying to install Amber 9 on a 64 bit Xeon machine with
> RHEL 5.1 as the operating system. I used gcc [version: gcc (GCC) 4.1.2
> 20070626 (Red Hat 4.1.2-14) ] and g95 [version: G95 (GCC 4.0.3 (g95
> 0.90!) Jul 27 2006)] (./configure g95) and the compilation was
> succesful, and I did not notice any errors along the way. But, when I
> ran the test suite, several of the tests gave results that were
> significantly far from what was expected, and in particular the ptraj
> module completely crashed with the following errors:
> ----------------------------------------
> # make test.ptraj
> cd ptraj_matrix; ./Run.matrix
>
> ptraj: analyze fluctuation matrices
> Segmentation fault
> ./Run.matrix: Program error
>
> make: *** [test.ptraj] Error 1
> ----------------------------------------
>
> Does anyone have any ideas what the problem might be? I am guessing
> that something is wrong with the fortran compilation? Does anyone know
> if some other version of the compilers would be more useful?
>
> Looking through the archives I noticed that a 32 bit installation is
> sometimes recommended. What options should I change in the config.h
> file or Makefiles to attempt that?
>
> Also, looking through the archives I noticed that the intel compiler
> was highly recommended. I tried installing Amber9 with gcc as the c
> compiler and the trial version of the Intel fortran compiler [version
> 10.1.012] and got the following output:
>
> -----------------------------------------------------------------------------
> # export MKL_HOME=/opt/intel/mkl/10.0.1.014
> # ./configure -static ifort_x86_64
> AMBERHOME is set to /usr/local/Amber/amber9
> Setting up Amber configuration file for architecture: ifort_x86_64
> Using parallel communications library: none
> MKL_HOME is set to /opt/intel/mkl/10.0.1.014
> Using MKL libraries from /opt/intel/mkl/10.0.1.014/lib/em64t
>
> The configuration file, config.h, was successfully created.
>
> [root_at_auzanneau8 src]# make serial
> Starting installation of Amber9 (serial) at Tue Mar 18 11:34:50 EDT 2008.
> cd lib; make install
> make[1]: Entering directory `/usr/local/Amber/amber9/src/lib'
> cpp -traditional -P -DMKL new2oldparm.f > _new2oldparm.f
> ifort -c -w95 -mp1 -O0 -o new2oldparm.o _new2oldparm.f
> ifort: error #10149: option '-cxxlib-gcc' not supported with the
> current installed GCC
> make[1]: *** [new2oldparm.o] Error 1
> make[1]: Leaving directory `/usr/local/Amber/amber9/src/lib'
> make: *** [serial] Error 2
> [root_at_auzanneau8 src]#
> -----------------------------------------------------------------------------
>
> I also tried using both Intel c and fortran compilers version 10.1.012
> and got the same error.
>
> Does anyone know what version of Intel compilers might be compatible
> with gcc or what version of gcc is likely to work with the Intel 10
> compiler?
>
> Or am I missing something else?
>
>
> Thanks alot!
>
> Trudy
>
>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
>
>
>         Cenk Andac, M.S., Ph.D. Student
>
> School of Pharmacy at
> Gazi University-Ankara Turkiye
>
>
>   Address: Bandirma Sok. No:6
>
>   Etiler, Ankara, 06330 Turkey
>
>   Cell: +90-(536)-4813012
>
>   E-Mail:cenk_andac_at_yahoo.com
>
>
>
>
>
> ---------------------------------
>   Be a better friend, newshound, and know-it-all with Yahoo! Mobile.  
>  Try it now.
>
>
> ---------------------------------
>   Looking for last minute shopping deals? Find them fast with Yahoo! Search.
>
>
> ---------------------------------
> Never miss a thing.   Make Yahoo your homepage.

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu