home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!julienas!chorus!octave.chorus.fr!jloup
- From: jloup@chorus.fr (Jean-loup Gailly)
- Newsgroups: comp.compression
- Subject: Re: How much memory does MW compression really need?
- Message-ID: <1420@chorus.chorus.fr>
- Date: 14 Dec 92 16:04:45 GMT
- References: <1992Dec13.195334.5652@plex-1.ann-arbor.mi.us>
- Sender: jloup@chorus.chorus.fr
- Reply-To: jloup@chorus.fr (Jean-loup Gailly)
- Lines: 31
-
- David West <dhw@plex-1.ann-arbor.mi.us> writes:
-
- > I've compiled Dan Bernstein's implementation of Miller-Wegman compression
- > ("sqz" from comp.sources.somethingorother).
-
- comp.sources.unix, Volume 21, Issue 1, 6 Feb 90.
-
- > The included comments say that it's "even more of a memory hog than
- > Compress". Well, compress runs in 400K, but sqz (with no other tasks
- > running) thrashes and then dies on my Unix box (6MB real memory, 5MB
- > swap space) trying to compress a 100K file. Anyone know how much
- > memory is comfortable for it? Or is something else the problem?
-
- On the file book2 from the Calgary corpus, squeeze requires 5M of
- virtual memory and uses about 4M of physical memory (obtained with
- "ps -v" on a SparcStation SLC with 8M of physical memory). By comparison,
- zip 1.9 uses about 0.5M of physical memory, runs 4 to 7 times as fast
- (elapsed time on an idle machine) and compresses better than squeeze:
-
- original compress squeeze zip -5 zip -1
-
- size 610,856 250,759 230,862 206,771 217,341
- user 4.840u 18.350u 13.030u 8.040u
- system 0.600s 4.140s 0.500s 0.390s
- elapsed 0:05.93e 1:08.69e 0:15.08e 0:09.39e
-
- Because of the higher memory usage, squeeze pages much more. This
- explains the higher ratio elapsed time/cpu time.
-
- Jean-loup Gailly
- jloup@chorus.fr
-