From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-447148-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 27656 invoked by alias); 22 Mar 2014 00:48:45 -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 27455 invoked by uid 48); 22 Mar 2014 00:48:41 -0000
From: "pinskia at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/60604] GCC incorrectly compiles s_csinh function on MIPS32 (32bit fp)
Date: Sat, 22 Mar 2014 00:48:00 -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: 4.9.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: pinskia at gcc dot gnu.org
X-Bugzilla-Status: UNCONFIRMED
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: cf_gcctarget short_desc
Message-ID: <bug-60604-4-2zFUDhlAxJ@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-60604-4@http.gcc.gnu.org/bugzilla/>
References: <bug-60604-4@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: 2014-03/txt/msg02017.txt.bz2

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60604

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|mips*-*-*                   |mips*-*-* (o32/eabi32)
            Summary|GCC incorrectly compiles    |GCC incorrectly compiles
                   |s_csinh function on MIPS    |s_csinh function on MIPS32
                   |                            |(32bit fp)

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The code generation looks incorrect to me:

    mfc1    $3,$f12
    mfc1    $2,$f12
    move    $17,$3
    jal    myclassify
    ext    $16,$2,0,31


Notice how $2/$3 are moving from the same register, one of them should have
been $f13.


The problem is due to paired floating point registers are swapped for
big-endian.
/* Paired FPRs are always ordered little-endian.  */

So when the register allocator is figuring out which register to copy from for
the high subreg of the DF mode:
(insn 23 22 24 (set (subreg:SI (reg:DF 200 [ D.2940 ]) 0)
        (and:SI (subreg:SI (reg:DF 194 [ D.2940 ]) 0)
            (const_int 2147483647 [0x7fffffff]))) t77.c:25 -1
     (nil))

It decides that is the same as the register which is incorrect for pair float.

This is what the register allocator produces:

(insn 110 8 23 2 (set (reg:SI 2 $2)
        (reg:SI 44 $f12)) t77.c:25 302 {*movsi_internal}
     (nil))
(insn 23 110 111 2 (set (reg:SI 16 $16 [ D.2940 ])
        (and:SI (reg:SI 2 $2)
            (const_int 2147483647 [0x7fffffff]))) t77.c:25 157 {*andsi3}
     (nil))

This is incorrect as we should be using $f13 or the full DI mode instead.