From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,8f8cea8602e61aba
X-Google-Attributes: gid103376,public
From: kenner@lab.ultra.nyu.edu (Richard Kenner)
Subject: Re: The Red Language
Date: 1997/09/22
Message-ID: <604jv0$rpo$1@news.nyu.edu>#1/1
X-Deja-AN: 274512969
References: <dewar.874664808@merv> <EGrv5y.C2D@world.std.com>
 <dewar.874850594@merv>
Organization: New York University Ultracomputer Research Lab
Newsgroups: comp.lang.ada
Date: 1997-09-22T00:00:00+00:00
List-Id: <comp.lang.ada>


In article <dewar.874850594@merv> dewar@merv.cs.nyu.edu (Robert Dewar) writes:
>The rub is "full", just how full must a traversal be to be considered
>a pass. For example, in GNAT, there is a circuit at the end of the
>parser to fill in declarations for labels. Is this a pass? I would
>say no, but it is open to argument. As soon as you have random access
>to the representation, the definition is no longer clear (and is
>incidentally, not particularly useful -- in the old days there was
>far more connection between the notions of efficiency in compilation
>time and number of passes than there is today).

In GCC, trying to define a "pass" is even harder.

Consider something like this, from combine.c:

  for (insn = f, i = 0; insn; insn = NEXT_INSN (insn))
    if (INSN_UID (insn) > i)
      i = INSN_UID (insn);

This loop is finding the highest value of INSN_UID for use in an array
allocation.  But this does scan all instructions in the subprogram
being compiled, so it might, by some definitions, be considered a
"pass".  But there are dozens of things like this all over the
compiler, to scan for one thing or another.