David, Thanks for your responce.
1) The cluster has P4s (uname -m returns i686).
2) Compiling with icc 9.1 and with -m32 removed from config.h gave no
errors during "make test.serial", however, I do remember errors using gcc.
3) m32 is not in the gcc 2.96 man page so it may not be a ligitamite
flag in gcc 2.96.
Since amber 9.0 seems to compile and run qith icc 9.1 on our system
we'll go with this.
Thanks,
Don
David A. Case wrote:
>On Mon, May 22, 2006, Don Harden wrote:
>
>
>
>>However the compilation dies with gcc -m32 -c -D_FILE_OFFSET_BITS=64
>>-D_LARGEFILE_SOURCE -O2 -o mmtsb_client.o mmtsb_client.c
>>-----------------------------------
>>cc1: Invalid option `32'
>>
>>
>
>
>
>>Removing the -m32 flag at CC and LOADCC in config.h allows amber to
>>compile, but the test suit dies.
>>
>>
>
>Some basic questions:
>
>1. What does "uname -m" return:, i.e. what sort of CPU do you have?
>
>2. What part of the test suite gives problems? Note that mmtsb_client.c is
>just a dummy routine, so it's really frustrating (for both users and us!)
>that C-compiler problems are inhibiting compilation of sander.
>
>If you also set -DF90_TIMER in AMBERBUILDFLAGS, there will be no C code
>at all in sander. This should isolate Fortran compiler problems from gcc
>problems. So (if it is indeed sander that is failing in the test suite) you
>might try that option.
>
>3. Does the man page for your gcc installation indicate that "-m32" is
>supposed to be a valid option? It certainly is true that we have not
>tested things with gcc 2.96 (and that particular version always had some
>problems, compared to 2.95 or 3.x or 4.x), but these messages are mysterious.
>
>...regards...dac
>-----------------------------------------------------------------------
>The AMBER Mail Reflector
>To post, send mail to amber.scripps.edu
>To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
>
--
Don Harden harden.Gsu.EDU
Department of Chemistry 564 NSC
Georgia State University ph: (404) 651-3580
Atlanta, Ga. 30302 fax: (404) 651-1416
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed May 24 2006 - 06:07:16 PDT
