Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!oakhill!dover!digital!chen@digital.sps.MOT.COM
From: chen@digital.sps.MOT.COM (Jinfu Chen)
Newsgroups: comp.sys.atari.st
Subject: lharc/arc/zoo/compress
Message-ID: <47f0d326.81da@digital.sps.mot.com>
Date: 10 Jan 90 00:31:45 GMT
Sender: chen@digital.sps.mot.com
Reply-To: chen@digital.sps.MOT.COM (Jinfu Chen)
Organization: Motorola, Inc. Logic IC Div, Mesa, AZ
Lines: 64

I did a quick and dirty comparision on four compression programs in an UNIX
box.And here is the result:


 Program  Size  Cmp Size Ratio Comp. Time Decmp. Time Comp. Speed Decomp. Speed
          byte    byte     %     second     second    byte/second  byte/second

(Text)

   lharc 961244 253799   73.6    353.8       63.8        2717         15067
compress 961244 302096   68.6     29.4       22.6       32695         42533
     zoo 961244 368158   61.7     85.9       48.2       11190         19943
     arc 961244 425193   55.8    163.4      111.0        5883          8660

(Binary)
                                            
   lharc 293489 130670   55.5    101.7       32.3        2886          9086
compress 293489 167001   43.1     13.5        9.2       21740         31901
     zoo 293489 170255   42.0     33.8       17.2        8683         17063
     arc 293489 180114   38.6     55.7       37.8        5269          7764

Text file: xlib document (nroff output)      
Binary file: mush (apollo m68k coff) 

Program versions:        

compress 5.1   
lharc V0.03 (Beta)       
arc 5.21
zoo 2.01

Observations:

o	In terms of compresion size, lharc ranks on top, with 5-12% better than
	compress,  12-13% better than zoo, and 17-18% better than arc(5.21).

o	In terms of speed in compression mode, compress rates much higher than the
	rest, in the order of 2.5-3 times faster than zoo, 4-6 times faster than 
	arc (5.21), and 8-12 times faster than lharc.

o	In decompression mode, the speed differences are less than the ones in 
	compression mode. Compress still rates the fastest, with 2 times faster than
	zoo, 2.5-3 times faster than lharc (not bad), and 4-5 times faster than
	arc (5.21).

There's no doubt people who use online services love `lharc', in which
transfer time is their main concern, giving the fact that `lharc' provides
over 10% better compression ratio than other methods. While other situation
when storage space and process time are to be balanced, `compress' probably is
the choice.

I read from the README file in arc 6.0 (ST version) that 6.0's speed is quite
fast comparing to 5.21. However, the compression ratio changes very little.
I doubt even arc 6.0 can beat other three programs in both speed and
compression.

Of course there are other issues involved when comparing these programs, such
as feature of including comments, capability of including multiple files
and/or directories. For now I just simply ignore them.

In case anyone interested, the tests were run on an Apollo workstation DN3000
(12 Mhz 68020/68881) with 8M memory (OS SR10.2), which is approximately a 1.9
mips machine (comparing to the 0.8-0.9 mips 8Mhz 68000 ST). Run times are
reported in user's time.