From: Niall (niall at domain magicgoeshere.com)
Date: Wed 17 May 2000 - 17:36:18 IST
On Wed, May 17, 2000 at 12:22:20PM +0100, Kenn Humborg wrote:
> > the %MEM fields that top displays I get 191.7 - I presume that's the
> > kernel's memory overcommit coming into effect :-)
>
> Probably more due to shared memory. All the code in shared
> libraries and executables is only loaded once into RAM
> and shared between processes. So if you have two bash processes
> each looks like this (numbers are fictional):
>
> Process 1: code = 1000k data = 200k
> Process 2: code = 1000k data = 300k
>
> In top, these will show up with RSS values of 1.2M and 1.3M.
> So, if you total these, you naively assume that they take
> up 2.5MB of RAM, even though they only take up 1.5M because
> the code portions are shared.
Yes, I did of course have quite some shared memory, in particular with the
soffice.bin instances, but if you look back at my output from free, there
was about 19M of shared memory, and about 29M free (including
buffers/cached) so there's still a big discrepancy there between total RSS
and physical memory, a chunk of which I'm presuming is memory overcommit - I
can't think of any other explanation for it.
Regards,
Niall
This archive was generated by hypermail 2.1.6 : Thu 06 Feb 2003 - 13:06:12 GMT