:I want to see at least a basic measurements first. This is very likely
:to increase both the runtime and memory foot print of programs and the
:argument "OpenBSD did it" is not a good justification for both.
:
:Joerg
Well, you are welcome to make such measurements but I don't see anything
in the patch that would increase memory footprints, since the old code
was sbrk()ing memory in multiples of PAGE_SIZE and the new code
effectively does the same thing, just with mmap() instead of sbrk().
There is some potential for a more fragmented VM map for processes
which mix malloc()'s and their own file mmap()'s, but that's about it.
Sans a process doing its own mmap()ing, the effect on the VM map will
be the same.
I will do a /usr/bin/time -l on a buildworld before and after the patch
to make sure that there are no severe differences. I doubt I'll see
any.
-Matt
Matthew Dillon
<[EMAIL PROTECTED]>
