c/5707: i386 compiler can't build libgcc2.c with -msoft-float

c/5707: i386 compiler can't build libgcc2.c with -msoft-float

[eric.norum]

>Number:         5707
>Category:       c
>Synopsis:       i386 compiler can't build libgcc2.c with -msoft-float
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Feb 16 07:46:01 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Eric Norum <eric.norum@usask.ca>
>Release:        2002-02-14 snapshot
>Organization:
>Environment:
Redhat 7.2 linux
>Description:
Here's what I can cut the failing program down to (this is part of libgcc2.c).
=================================================================
typedef float DFtype __attribute__ ((mode (DF)));
typedef int DItype __attribute__ ((mode (DI)));
typedef int SItype __attribute__ ((mode (SI)));
typedef unsigned int USItype __attribute__ ((mode (SI)));
typedef unsigned int UDItype __attribute__ ((mode (DI)));
DFtype
__floatdidf (DItype u)
{
  DFtype d;

  d = (SItype) (u >> (sizeof (SItype) * 8));
  d *= (((UDItype) 1) << ((sizeof (SItype) * 8) / 2));
  d *= (((UDItype) 1) << ((sizeof (SItype) * 8) / 2));
  d += (USItype) (u & ((((UDItype) 1) << (sizeof (SItype) * 8)) - 1));

  return d;
}
=================================================================

Compiler details:
../gcc/configure --enable-languages="c,c++"  --prefix=/tmp/cjnk

source@norum1 1633> /tmp/cjnk/bin/gcc --version
gcc (GCC) 3.1 20020214 (experimental)
Copyright (C) 2002 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
=================================================================


Here's what happens when it's compiled with optimization turned on:
source@norum1 1635> /tmp/cjnk/bin/gcc -c -O2 -msoft-float b.c
b.c: In function `__floatdidf':
b.c:17: unrecognizable insn:
(insn 13 12 16 (set (reg/v:DF 60)
        (float:DF (subreg:SI (reg:DI 61) 0))) -1 (insn_list 12 (nil))
    (expr_list:REG_DEAD (reg:DI 61)
        (nil)))
b.c:17: Internal compiler error in extract_insn, at recog.c:2129
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

=================================================================


And here's what happens when it's compiled with optimization turned off:
source@norum1 1636> /tmp/cjnk/bin/gcc -c -msoft-float b.c
b.c: In function `__floatdidf':
b.c:17: Internal compiler error in instantiate_virtual_regs_1, at function.c:3880
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
=================================================================
>How-To-Repeat:
Try to build libgcc2.c with -msoft-float
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:

Re: c/5707: i386 compiler can't build libgcc2.c with -msoft-float

[Richard Henderson]

The following reply was made to PR c/5707; it has been noted by GNATS.

From: Richard Henderson <rth@redhat.com>
To: Joel Sherrill <joel.sherrill@OARcorp.com>
Cc: gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org, eric.norum@usask.ca,
   gcc-bugs@gcc.gnu.org, Jan Hubicka <jh@suse.cz>
Subject: Re: c/5707: i386 compiler can't build libgcc2.c with -msoft-float
Date: Tue, 5 Mar 2002 14:10:10 -0800

 On Tue, Mar 05, 2002 at 06:15:19AM -0600, Joel Sherrill wrote:
 > Richard/Jan could you try this out and take a stab at fixing it?  This
 > is a real pain for the embedded community using i386dx/ex and 486sx 
 > derived CPUs.  
 
 See if this is all you need.
 
 
 r~
 
 
 
 Index: config/i386/i386.md
 ===================================================================
 RCS file: /cvs/gcc/gcc/gcc/config/i386/i386.md,v
 retrieving revision 1.339
 diff -c -p -d -u -r1.339 i386.md
 --- i386.md	2002/02/21 21:16:18	1.339
 +++ i386.md	2002/03/05 22:26:35
 @@ -5136,7 +5136,7 @@
  (define_expand "floatsidf2"
    [(set (match_operand:DF 0 "register_operand" "")
  	(float:DF (match_operand:SI 1 "nonimmediate_operand" "")))]
 -  ""
 +  "TARGET_80387 || TARGET_SSE2"
    "")
  
  (define_insn "*floatsidf2_i387"

Re: c/5707: i386 compiler can't build libgcc2.c with -msoft-float

[Joel Sherrill]

The following reply was made to PR c/5707; it has been noted by GNATS.

From: Joel Sherrill <joel.sherrill@OARcorp.com>
To: joel.sherrill@oarcorp.com, gcc-gnats@gcc.gnu.org,
 	gcc-prs@gcc.gnu.org, eric.norum@usask.ca, gcc-bugs@gcc.gnu.org,
 	Jan Hubicka <jh@suse.cz>, rth@redhat.com
Cc:  
Subject: Re: c/5707: i386 compiler can't build libgcc2.c with -msoft-float
Date: Tue, 05 Mar 2002 06:15:19 -0600

 Eric sent me email privately confirming that PR5707 is present on 
 i386-linux-gcc as well as on i386-rtems-gcc.  It's the -msoft-float that 
 causes the problem.  When we tried adding -msoft-float as a multilib
 to the i386-elf target in the past, it has broken as well.  This is 
 on the main line CVS and a regression from 3.0.4.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=5707
 
 Richard/Jan could you try this out and take a stab at fixing it?  This
 is a real pain for the embedded community using i386dx/ex and 486sx 
 derived CPUs.  
 
 Thanks.
 
 -- 
 Joel Sherrill, Ph.D.             Director of Research & Development
 joel@OARcorp.com                 On-Line Applications Research
 Ask me about RTEMS: a free RTOS  Huntsville AL 35805
    Support Available             (256) 722-9985