Some basic MM info:

When a process fork()'s (vfork()'s or whatever)
it's page tables are copied and the pages of the new process are
marked as COW (Copy On Write). If the new process writes to any
of it's pages the kernel handles a page fault, copies the page and
maps it back in. The pratical effect being that all memory which
can be shared, is shared and fork() is fast.

Now - if you malloc 10MB of memory and fork() you will show up as
using 20MB of memory, even thou 10MB is actually used. Since fred
has so many threads, this could be really screwing the memory
use figures.

AGL

-- 
No. The bug is *not* in gcc.

PGP signature

Reply via email to