From: tkane01@unix.cc.emory.edu (Terrence O Kane)
Message-Id: <9501242217.AA21979@larry.cc.emory.edu>
Subject: Re: Beta test for DOS
To: vern@ee.lbl.gov (Vern Paxson)
Date: Tue, 24 Jan 1995 17:17:38 -0500 (EST)
In-Reply-To: <199501232138.NAA11430@daffy.ee.lbl.gov> from "Vern Paxson" at Jan 23, 95 01:38:02 pm
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 5792      

Vern,

I've made flex 2.5.0.6 successfully with no changes to the source
code at all.  I'm including the Borland 4.02 makefile and config.h
at the end of this message.

When you're ready for release, I'll be happy to take care of getting
the appropriate archive files up to Simtel if you wish.

I have not used this version for any of my "work-related" scanners
yet, but have run the fastwc benchmark.  The compiles were for large
memory model and optimization for fastest possible code.  The test
machine was a Pentium-90 (hey! timing output was integer!) with
enhanced IDE on a PCI bus and no file system caching.  I ran the
test on two different input files.

(Times are in seconds.)

The first input was a typical postscript file concatenated 10 times;
by typical, I mean that there were relatively few spaces, lots of lines
with no space, using lots of parentheses.

	lines  words  characters
	91200  356260 5889240

	mywc 	 8.89
	wc1s 	15.22		default
	wc1 	10.82		-Cf
	wc2 	10.16		-Cf
	wc3 	 9.17		-Cf
	wc4 	 9.22		-Cf
	wc5 	10.98		-Cf

The second test run was on a file that consisted of 20 concatenations
of 'misc/flex.man'.

	lines  words  characters
	69960   305140  2399960

	mywc 	4.01
	wc1s 	6.87
	wc1 	5.32
	wc2 	4.95
	wc3 	4.12
	wc4 	4.12
	wc5 	5.05

[[Makefile and config.h moved to separate files -VP]]