LW| I decided to compress the latest FidoNet nodelist (NODELIST.299).
| Using JAR, it took roughly 29 minutes on my 386SX/16, and the
Yikes! You sure you had the turbo switch turned on?
LW| archive size was 983207 bytes (using maximum compression).
| Then I checked with ARJ v.2.50. It took less than 4 minutes, and the
| archive size was 984399. On the same machine.
Interesting... I tried a few tests on that same file. Here are the
results:
Program Time Size
JAR32 20.9 924,034 (used 6 megs of RAM)
JAR32 -m4 42.5 912,346 (used 6 megs of RAM)
JAR 26.0 978,872 (used 0.5 megs of EMS)
JAR -m4 49.4 967,433 (used 0.5 megs of EMS)
JAR 30.9 962,586 (used 4 megs of EMS and 1 meg of XMS)
JAR -m4 57.1 952,999 (used 4 megs of EMS and 1 meg of XMS)
ARJ 15.2 974,621
ARJ -jm 16.3 974,280
I'm using a Pentium 133 w/32 megs of RAM running Windows 95, and I used
the JAR.CFG supplied in the distribution archive. I used the 2.55b
version of ARJ.
The first set of JAR (16-bit) tests were made with my usual DOS session
settings, which allow 1 meg of EMS and 1 meg of XMS. For the second
set, I gave it unlimited amounts of both. With these settings, the MEM
command reports 16 megs of EMS free and approximately 29 megs of XMS
free.
LW| Actually I don't think it's really worth waiting another 25 minutes
| to get a saving of 1192 bytes ;-)
I agree completely! It's strange that in your test, JAR took about 7.25
times as long as ARJ, while in my equivalent tests, it took about 3 to
3.5 times as long (comparing the two JAR -m4 tests to ARJ -jm).
It's also interesting that with both programs using their default
compression, ARJ actually compressed better than JAR when JAR was using
0.5 megs of EMS. Giving JAR more RAM changed that, but it looks like
you have to use JAR32 to gain any real advantage in compression.
Of course, it's tough to draw any conclusions form a simple test like
this. We'll have to see what happens when the experts like Jeff
Gilchrist and Joe Negron give JAR a real workout!
---
* OLXWin 1.00b * Internet: 71534.444@compuserve.com
--- QScan/PCB v1.17b / 01-0035
---------------
* Origin: Agora Online -- Bayonne, NJ 201-437-4355 v.34 (1:2630/312)
|