Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-lcc!ames!oliveb!felix!ccicpg!cci632!ritcv!rochester!pt.cs.cmu.edu!sei!sei.cmu.edu!firth
From: firth@sei.cmu.edu (Robert Firth)
Newsgroups: comp.arch
Subject: Comparison Data
Message-ID: <1290@aw.sei.cmu.edu>
Date: Thu, 7-May-87 15:46:48 EDT
Article-I.D.: aw.1290
Posted: Thu May  7 15:46:48 1987
Date-Received: Sat, 9-May-87 18:40:42 EDT
Sender: netnews@sei.cmu.edu
Lines: 40

Postscript to previous post: raw data
gathered from systems code.

Comparison	%

var = 0		18
var >=0		12
var > 0		 0

var = const	31
var >(=)const	18

var = var	19
var > var	 2
var >=var	 1

Each comparison counts both the test and the converse
test, so 'v>0' includes 'v<=0'.  The test 'v>const'
includes all 4 forms, since we can always transform
'v>k' into 'v>=(k-1)'.  Basically 'var' means "anything
not known at compile time"; I didn't check for things like

	IF x > y+1 ...

that can be transformed into IF x>=y.

Gathered from ~12000 lines of BCPL code.  Note two
language-dependent sources of bias

(a) strings have a length count not a NUL terminator,
    so we have 'i<=length' where C would have 'ch=0'

(b) many library calls return a negative error code
    tested for by '<0' comparison

Booleans are represented by (FALSE=>0, TRUE=>allones),
and always tested by treating 0 as FALSE and everything
else as TRUE.

Hope this is informative.