Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: $Revision: 1.6.2.16 $; site ima.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!think!ima!johnl
From: johnl@ima.UUCP
Newsgroups: net.lang
Subject: Re: compiling nested conditionals
Message-ID: <39700001@ima.UUCP>
Date: Tue, 18-Jun-85 14:16:00 EDT
Article-I.D.: ima.39700001
Posted: Tue Jun 18 14:16:00 1985
Date-Received: Fri, 21-Jun-85 01:19:24 EDT
References: <328@sbcs.UUCP>
Lines: 23
Nf-ID: #R:sbcs:-32800:ima:39700001:000:746
Nf-From: ima!johnl    Jun 18 14:16:00 1985


> My solution has been to generate the label of the ultimate branch target
> and pass it down during syntax-directed translation, so that goto-chains
> are never generated.

Both the Ritchie and PCC C compilers and their f77 and Pascal 
cousins use this approach for compliling complicated conditional 
expressions.  As noted, it works very well.  It's particularly 
useful for things like Ratfor output, since Ratfor tends to translate this:

    if (expression) stuff

into this:

    if(.not.(expression)) goto 12345
    stuff
12345 continue

Even if the conditional expression is pretty complicated, the compilers
unscramble them well enough that there is no evidence in the runtime code
that the scrambling happened.

John Levine, ima!johnl