From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-464436-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 7157 invoked by alias); 19 Oct 2014 02:34:45 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Received: (qmail 7076 invoked by uid 48); 19 Oct 2014 02:34:33 -0000
From: "kkojima at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/55212] [SH] Switch to LRA
Date: Sun, 19 Oct 2014 02:34:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: unknown
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: kkojima at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-55212-4-nhbcYKN5TS@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-55212-4@http.gcc.gnu.org/bugzilla/>
References: <bug-55212-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-10/txt/msg01457.txt.bz2

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55212

--- Comment #69 from Kazumoto Kojima <kkojima at gcc dot gnu.org> ---
The patch in c#57 disables memory equiv substitution for the memory
with base+index and base+display addressing.

static bool
sh_cannot_substitute_equiv_p (rtx subst)
{
  if (TARGET_SHMEDIA)
    return false;

  if (GET_CODE (subst) == SUBREG)
    subst = SUBREG_REG (subst);
  if (MEM_P (subst) && GET_CODE (XEXP (subst, 0)) == PLUS)
    return true;

  return false;
}

A bit surprisingly, disabling all memory equiv substitution wins
CSiBE tests.  With

static bool
sh_cannot_substitute_equiv_p (rtx subst)
{
  if (TARGET_SHMEDIA)
    return false;

  return true;
}

the code size regression for CSiBE from non LRA is reduced to 0.59%.
Looking at the improved cases, the extra save/restore insns to/from
stack disappear.  I guess that SH has not enough numbers of the hard
registers to make the equiv substitution win in the size and the speed
on average working sets.  It looks the pseudos produced to hold
the equiv values can't get hard registers for bad cases and end up
memory save/restore insns which make the code worse.