Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!cs.utexas.edu!cse!utafll!john
From: john@utafll.uta.edu (John Baima)
Newsgroups: comp.lang.smalltalk
Subject: Richards benchmark an a DOS machine
Message-ID: <1991Jun10.144219.16193@cse.uta.edu>
Date: 10 Jun 91 14:42:19 GMT
Sender: usenet@cse.uta.edu (USENET Dummy account for GNUS and TheNews)
Organization: Computer Science Engineering Univ. of Texas at Arlington
Lines: 17

I ran the Richards benchmark on a 16Mhz 386sx with Borland's C++ 2.0
and Smalltalk-80 rel 4. The BC++ average was about 8.1 seconds (run
under DOS). The ST-80 average was about 29.9 seconds (run under
Windows 3.0). The Smalltalk times showed significantly more variation
(6 seconds) than the BC++ times. I suppose that this is due to Windows
fooling around in the background. Thus, the Smalltalk times are about
3.7 times slower. However the Smalltalk program was run under Windows
and I expect that that slows things down some even without writing to
the screen.

I think that the times for ST-80 ver 2.4 will be a good deal slower
than ver 4.0 and should not be used as a true benchmark.

I'd be interested in other benchmarks and other 'Richards' results.
-- 
John Baima
john@utafll.uta.edu