public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "kazu at cs dot umass dot edu" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug optimization/15349] New: [tree-ssa] Merge two phi nodes.
Date: Sun, 09 May 2004 19:48:00 -0000	[thread overview]
Message-ID: <20040509194800.15349.kazu@cs.umass.edu> (raw)

void bar (void);

int
foo (int a, int b)
{
  int t;

  if (b)
    {
      if (a)
	t = 3;
      else
	t = 5;

      if (t)
	a++;
    }
  else
    t = 7;

  return t - 1;
}

I get:

foo (a, b)
{
  int t;

<bb 0>:
  if (b_3 != 0) goto <L0>; else goto <L7>;

<L0>:;
  if (a_6 != 0) goto <L1>; else goto <L3>;

<L1>:;

  # t_1 = PHI <5(1), 3(2)>;
<L3>:;

  # t_2 = PHI <7(0), t_1(3)>;
<L7>:;
  return t_2 - 1;

}

Note that we can merge the first PHI into the second one like so:

  # t_2 = PHI <7(0), 5(1), 3(2)>;

by replacing goto <L1>; with goto <L7>;

Note that this optimization is pretty important because I see alias.c
containing the following:

<L191>:;

  # iftmp.1496_89 = PHI <0(139), 1(138)>;
<L192>:;
  iftmp.1493_304 = iftmp.1496_89 != 0;

  # iftmp.1493_88 = PHI <iftmp.1493_304(140), iftmp.1493_316(136)>;
<L193>:;
  if (iftmp.1493_88) goto <L194>; else goto <L202>;

Once PR 14797 is solved, the above code should become something like

<L191>:;

  # iftmp.1493_304 = PHI <0(139), 1(138)>;
<L192>:;

  # iftmp.1493_88 = PHI <iftmp.1493_304(140), iftmp.1493_316(136)>;
<L193>:;
  if (iftmp.1493_88) goto <L194>; else goto <L202>;

Once this PR is solved, we would have

<L192>:;

  # iftmp.1493_88 = PHI <0(139), 1(138), iftmp.1493_316(136)>;
<L193>:;
  if (iftmp.1493_88) goto <L194>; else goto <L202>;

Now it's trivial to redirect jumps from bb 139 and bb 138
to <L202> and <L194>, respectively, provided that nobody else
uses iftmp.1493_88.

-- 
           Summary: [tree-ssa] Merge two phi nodes.
           Product: gcc
           Version: tree-ssa
            Status: UNCONFIRMED
          Keywords: pessimizes-code
          Severity: enhancement
          Priority: P2
         Component: optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: kazu at cs dot umass dot edu
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15349


             reply	other threads:[~2004-05-09 19:48 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-05-09 19:48 kazu at cs dot umass dot edu [this message]
2004-05-09 20:36 ` [Bug optimization/15349] " pinskia at gcc dot gnu dot org
2004-05-21 13:32 ` [Bug tree-optimization/15349] " dnovillo at gcc dot gnu dot org
2004-05-25  0:11 ` pinskia at gcc dot gnu dot org
2004-09-14  2:55 ` pinskia at gcc dot gnu dot org
2004-09-14  3:53 ` kazu at cs dot umass dot edu
2004-09-19 19:03 ` kazu at cs dot umass dot edu
2004-09-20 20:08 ` kazu at cs dot umass dot edu
2004-12-22  4:59 ` kazu at cs dot umass dot edu
2005-01-20 19:21 ` cvs-commit at gcc dot gnu dot org
2005-01-20 19:49 ` kazu at cs dot umass dot edu
2005-01-20 21:10 ` pinskia at gcc dot gnu dot org

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20040509194800.15349.kazu@cs.umass.edu \
    --to=gcc-bugzilla@gcc.gnu.org \
    --cc=gcc-bugs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).