Re: [PATCH] proposed addition of Documentation/bigmem.txt

C. Scott Ananian (
Sun, 17 May 1998 20:25:49 -0400 (EDT)

On Sun, 17 May 1998 (Rogier Wolff) wrote:

> > On Sun, May 17, 1998 at 08:13:15AM -0400, Peeter Joot wrote:
> > > Are there some sort of disadvantages to machines with small
> > > memories that make this [default supports 2G RAM] undesirable?
> The default allows you to have 3G virtual adress space, with max 1G of
> physical RAM. The "hacked to allow 2G RAM" version only allows 2G of
> virtual adress space.

It should probably be clarified that this 'virtual address space' is per
process. The numbers quoted elsewhere were 3G physical, 1G virtual
memory. A balance like that would mean that you would need three separate
1G processes to use up all of your physical memory; since virtual <
physical, you couldn't have a single 3G process. The 2G/2G division is
more reasonable, I suppose. 'Swap space' is something entirely different,
because swap space is shared between all processes. If I've only 1G
virtual address space, then each process can use 1G max swap (if the
process was completely swapped out), but there's nothing to prevent me
from running multiple 1G processes, if I've got enough swap.

There *is* a bitwidth limit on swap, though; is it 32-bits/4G?
@ @
C. Scott Ananian: / Declare the Truth boldly and
Laboratory for Computer Science/Crypto / without hindrance.
Massachusetts Institute of Technology /META-PARRESIAS AKOLUTOS:Acts 28:31
-.-. .-.. .. ..-. ..-. --- .-. -.. ... -.-. --- - - .- -. .- -. .. .- -.
PGP key available via finger and from

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to