Kevin,

I'd rather say, it's perfectly OK to run 32bit software in a 64bit OS (on AMD64 or EM64T hardware, which behave the same for all practical purposes). Indeed my experience shows that 32bit software is almost as fast as its 64bit counterpart. Thus, there is no point in compiling 64bit binaries, unless the software/problem requires arrays of size > 2GB.

To install 32bit libraries (to be able to run 32bit software) together with the 64bit libraries on a 64bit OS is not a problem; they are usually distributed separately and end up in different directories (typically /usr/lib and /usr/lib64 , but that appears to depend on the distribution).

To _compile_ 32bit software on a 64bit OS is a different matter. There are three ways:
a) don't do it - just use a 32bit OS for compilation
b) start a shell with e.g. "linux32 bash", and -in that shell - compile with the -m32 flag in case of gcc/g77/gfortran. c) use the appropriate version of Intel compiler. Since release of 9.0, the 64 and 32bit version coexist on the same machine. The binaries produced also work very well on AMD hardware.

I run all crystallographic software including coot on a 64bit RHEL4 clone without any problems. My arsenal of commands to manage software consists mainly of "yum" and "rpm", helped by "file" and "ldd" to find out about binaries and libraries.

But it is of course true that if you anticipate that you'll not need more than 2 to 3 GB memory, there is no point in using a 64bit OS.

Kay

Kevin Cowtan wrote:
***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


Just in case it is not clear to everyone, it is perfectly OK to run a 32bit OS on a 64 bit machine. I'm using an Athlon X2 64 bit machine, but running a 32-bit Fedora on it. The performance difference is indistinguishable for typical tasks, and it means my binaries are usable everwhere else.

The downside is that I can't run apps which require more than 3Gb of memory. But I don't have that much memory anyway.

So in summary, by all means buy 64 bit machines. But unless you have a very specialised task to perform and are/have a guru to make it work for you, run it in 32 bit mode. You'll save yourself days of pointless effort.


Ethan Merritt wrote:

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


On Friday 26 May 2006 11:45 am, Christina Bourne wrote:

Prior to purchasing a computer with an AMD Turion 64
ML-40 processor, I was wondering if there were any
issues running Linux (SuSe or Ubantu) and packages
common to crystallography (ccp4, cns, o etc.)?



I don't know of any issues with the processor or the packages per se, but we continue to suffer problems with
64-bit code produced by gfortran.  To be specific, the
LAPACK library is in our experience unreliable when compiled with gfortran for AMD 64-bit.

It may be possible to work around the compiler bugs
with a proper set of compile options, but if so we have
not yet found them :-(   Several bugfixes relating to
this are apparently slated for the next release of
gfortran, but I don't know when they will percolate
into the various linux distributions.  I also don't
know whether they fix the specific problems we've
been having.





--
Kay Diederichs   http://strucbio.biologie.uni-konstanz.de
email: [EMAIL PROTECTED]  Tel +49 7531 88 4049 Fax 3183
Fachbereich Biologie, Universität Konstanz, Box M647, D-78457 Konstanz

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to