MOLPRO >2GB RAM on IA32/Linux?

[Nick Wilson]

Dear Jan,

An i4 molpro build will internally use a signed 32 bit integer to 
address memory which has a 2GB limit. If you want to address more than 
that then you'll have to make an i8 build using "configure -i8".

Best wishes,
Nick

Gershom (Jan M.L.) Martin wrote:
> Greetings:
> 
> SHORT VERSUON: Can MOLPRO use more than 2GB of RAM on a 
> P4 or Xeon machine running Linux?
> 
> LONG VERSION:
> 
> We are running MOLPRO on a cluster of dual Xeon machines running RedHat 8
> with updated kernels.
> 
> Even on the 4GB machines, the most memory I can get with a stock kernel
> is -m 253700000 (1936 MB) for a single-CPU job, and
> -m 194600000 (1485 MB) for a dual-CPU job.
> 
> Upon downloading kernel 2.4.24 from kernel.org and applying Andrea
> Arcangeli's "3.5GB user space" patch
> 
> http://ftp.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa1/00_3.5G-address-space-5                                               
> 
> I can now get -m 268200000 (2046 MB) in both single-and dual-CPU jobs.
> The largest continguous memory segment available is a bit more than 2.5 GB
> (due to the obnoxious Linux 7.3/8/9 habit of loading shared object libraries
> at the 1 GB mark --- you can see them by 'cat /proc/<processid>/maps) ---
> this can be increased a bit further by going to kernel 2.6 with Andrew
> Morton's "mm" patches (which include a separate 4GB/4GB user/kernel 
> space patch), and with a lot more doing (somehow forcing the shared object
> libraries to load at lower addresses --- ideas anybody?) this can probably
> be pushed over the 3 GB mark.
> 
> My question is now: is the P4/Linux version of MOLPRO capable at all of 
> handling memory spaces larger than 2GB? Or is this intrinsically off the
> menu?
> 
> Many thanks in advance
> Jan Martin
> --------------------------------------------------------------------
> Prof. Gershom (Jan M.L.) Martin      Computational Quantum Chemistry
>     Department of Organic Chemistry / Kimmelman Building, Room 252
>      Weizmann Institute of Science / IL-76100 Rechovot / ISRAEL
> Email:       comartin at wicc.weizmann.ac.il      Phone: +972(8)9342533
> Web site:  http://theochem.weizmann.ac.il/       FAX: +972(8)9344142
> --------------------------------------------------------------------
>