Re: AMBER: Problem with shared libraries:

From: David A. Case <>
Date: Sat, 6 May 2006 14:17:51 -0700

On Sat, May 06, 2006, Priti Hansia wrote:
> I am now trying to install Amber9 in our SGI-Altix cluster. Before
> starting the parallel installation I tried installing the serial one with
> g95 compiler. But I am getting the follwing error:
> [root.genomics src]# make serial
> Starting installation of Amber9 (serial) at Fri May 5 11:31:18 IST 2006.
> cd lib; make install
> make[1]: Entering directory `/software/amber9/src/lib'
> cpp -traditional -P -xassembler-with-cpp -DMKL new2oldparm.f >
> _new2oldparm.f
> g95 -c -O0 -fno-second-underscore -o new2oldparm.o _new2oldparm.f
> as: unrecognized option `-Qy'

Two points:

(1) As stated many times on the reflector in the past few weeks, if the
new2oldparm program will not compile, it is most likely a problem with your
compiler. Be sure that you can compile other fortran programs with your

(2) With a high-end machine like an SGI-altix, why do you want to use g95?
Do you not have access to ifort? It just seems strange to have such an
expensive and high-end parallel machine, and then couple it to a compiler
whose performance will be significantly slower than what should be
readily-available alternatives.

The g95 compiler is great if you are just learning amber, or are not planning
to do extensive simulations. Or if you don't wish to lay out the money for a
commerical compiler. But for extensive, parallel runs on high-end machines,
you should consider better alternatives.

(Also, I'm not sure anyone has ever tested the ia64 version of g95, for just
this reason. So you may have to be prepared to chart new ground here.)

...good luck...dac

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun May 07 2006 - 06:07:10 PDT
Custom Search