Hi!
On Wed, Aug 18, 1999 at 06:08:49PM +0200, Guido Flohr wrote:
> But I wouldn't be suprised if that penalty wasn't compensated by a
> performance goal in mark_region (less memory to mark)
Just a note in that context: I've sped up mark_region() (or more
precisely: mark_pages()) in mprot040.c a bit. Uses more memory now, but
it's faster. Don't expect miracles, though: A test loop on my Milan with
80 MB (without INITIALMEM) took 18s instead of 24 or so. The test was
just a "for i in /bin/*; do noop; done", where "noop" just immediately
Pterm0()ed. This is still very much, though even without memory
protection, it takes 10 seconds for "doing nothing" ;)
> Furthermore, most Unix software tends to malloc a lot of small and
> smallest blocks. I think that these tiny holes will not be wasted.
Don't forget that they malloc(), not sbrk() (where sbrk() may be
compared to Malloc(), though it's not exactly the same). Even Unix C
libraries use large blocks allocated from the system and then manage
them internally.
Ciao
Thomas
P.S.: Is it your intention that your mail was addressed to
"mint@stud.uni-sb.de" instead of "mint@fishpool.com"? It makes
replying a bit inconvenient ...
--
Thomas Binder (Gryf @ IRCNet) gryf@hrzpub.tu-darmstadt.de
PGP-key available on request! binder@rbg.informatik.tu-darmstadt.de
Vote against SPAM: http://www.politik-digital.de/spam/
Attachment:
pgpKjoDdDoE32.pgp
Description: PGP signature