Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[jakub]

Synopsis: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

State-Changed-From-To: feedback->closed
State-Changed-By: jakub
State-Changed-When: Mon Mar 18 15:37:59 2002
State-Changed-Why:
    Closing then.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=5968

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[Janis Johnson]

The following reply was made to PR optimization/5968; it has been noted by GNATS.

From: Janis Johnson <janis187@us.ibm.com>
To: jakub@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org,
        janis187@us.ibm.com, gcc-gnats@gcc.gnu.org
Cc:  
Subject: Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops
Date: Fri, 15 Mar 2002 11:00:58 -0800

 On Fri, Mar 15, 2002 at 01:44:31PM -0000, jakub@gcc.gnu.org wrote:
 > Synopsis: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops
 > 
 > State-Changed-From-To: open->feedback
 > State-Changed-By: jakub
 > State-Changed-When: Fri Mar 15 05:44:30 2002
 > State-Changed-Why:
 >     Are you sure you had my patch installed?
 >     Cannot reproduce it today at all and it really looks
 >     like 5891.
 > 
 > http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=5968
 
 OK, I'm embarrassed.  I updated my 3.1 tree yesterday and built and ran
 200.sixtrack and 254.gap; 200.sixtrack built and ran fine, so I assumed
 that I had picked up Jakub's patch, but it turns out I didn't.  Today I
 updated the tree again, verified that his patch is there, and both tests
 work fine.  Sorry for the confusion, PR 5968 is the same as PR 5891.
 
 I also saw runtime failures with 254.gap on i686-pc-linux.gnu, and with
 Jakub's fix for 5891 it now works fine.  I plan to continue building and
 running (with the small test input) SPEC CPU2000 benchmarks with a
 variety of optimization options and tracking down and reporting
 failures.
 
 Janis

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[jakub]

Synopsis: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

State-Changed-From-To: open->feedback
State-Changed-By: jakub
State-Changed-When: Fri Mar 15 05:44:30 2002
State-Changed-Why:
    Are you sure you had my patch installed?
    Cannot reproduce it today at all and it really looks
    like 5891.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=5968

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[jakub]

Synopsis: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

Responsible-Changed-From-To: unassigned->jakub
Responsible-Changed-By: jakub
Responsible-Changed-When: Thu Mar 14 19:20:30 2002
Responsible-Changed-Why:
    Mine.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=5968

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[Janis Johnson]

The following reply was made to PR optimization/5968; it has been noted by GNATS.

From: Janis Johnson <janis187@us.ibm.com>
To: law@redhat.com
Cc: janis187@us.ibm.com, gcc-gnats@gcc.gnu.org
Subject: Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops
Date: Thu, 14 Mar 2002 15:00:52 -0800

 On Thu, Mar 14, 2002 at 03:44:08PM -0700, law@redhat.com wrote:
 > In message <20020314222241.8465.qmail@sources.redhat.com>, janis187@us.ibm.com 
 > writes:
 >  > 
 >  > >Number:         5968
 >  > >Category:       optimization
 >  > >Synopsis:       flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loop
 >  > s
 >  > >Class:          ice-on-legal-code
 >  > >Submitter-Id:   net
 >  > >Arrival-Date:   Thu Mar 14 14:26:01 PST 2002
 >  > >Release:        gcc version 3.1 20020314 (prerelease)
 >  > >Environment:
 >  > Itanium, Red Hat Linux release 7.1.94 (Roswell)
 >  > >Description:
 >  > GCC 3.1 prerelease gets an ICE compiling 254.gap from SPEC
 >  > CPU2000 with "-O2 -funroll-loops" on ia64-unknown-linux-gnu.
 >  > Compiler output is:
 >  > 
 >  > bug.c: In function `f':
 >  > bug.c:49: Attempt to delete prologue/epilogue insn:
 >  > (insn 553 551 555 (set (reg:DI 333 ar.lc)
 >  >         (reg:DI 34 r35)) -1 (nil)
 >  >     (nil))
 >  > bug.c:49: Internal compiler error in propagate_one_insn, at flow.c:1573
 > There's a reasonable chance this patch from Jakub will fix your problem.
 > It was approved by Richard this afternoon.
 > 
 > > 	PR optimization/5891
 > > 	* unroll.c (copy_loop_body) [CALL_INSN]: Copy SIBLING_CALL_P flag.
 
 I should have mentioned in the PR that although this appears to be the
 same as the bug described in PR 5891, it occurs even with Jakub's fix
 for 5891.
 
 Janis

Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[law]

The following reply was made to PR optimization/5968; it has been noted by GNATS.

From: law@redhat.com
To: janis187@us.ibm.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 
 -funroll-loops
Date: Thu, 14 Mar 2002 15:44:08 -0700

 In message <20020314222241.8465.qmail@sources.redhat.com>, janis187@us.ibm.com 
 writes:
  > 
  > >Number:         5968
  > >Category:       optimization
  > >Synopsis:       flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loop
  > s
  > >Confidential:   no
  > >Severity:       serious
  > >Priority:       medium
  > >Responsible:    unassigned
  > >State:          open
  > >Class:          ice-on-legal-code
  > >Submitter-Id:   net
  > >Arrival-Date:   Thu Mar 14 14:26:01 PST 2002
  > >Closed-Date:
  > >Last-Modified:
  > >Originator:     Janis Johnson
  > >Release:        gcc version 3.1 20020314 (prerelease)
  > >Organization:
  > >Environment:
  > Itanium, Red Hat Linux release 7.1.94 (Roswell)
  > >Description:
  > GCC 3.1 prerelease gets an ICE compiling 254.gap from SPEC
  > CPU2000 with "-O2 -funroll-loops" on ia64-unknown-linux-gnu.
  > Compiler output is:
  > 
  > bug.c: In function `f':
  > bug.c:49: Attempt to delete prologue/epilogue insn:
  > (insn 553 551 555 (set (reg:DI 333 ar.lc)
  >         (reg:DI 34 r35)) -1 (nil)
  >     (nil))
  > bug.c:49: Internal compiler error in propagate_one_insn, at flow.c:1573
 There's a reasonable chance this patch from Jakub will fix your problem.
 It was approved by Richard this afternoon.
 
 > 	PR optimization/5891
 > 	* unroll.c (copy_loop_body) [CALL_INSN]: Copy SIBLING_CALL_P flag.
 
 jeff
 

optimization/5968: flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops

[janis187]

>Number:         5968
>Category:       optimization
>Synopsis:       flow.c ICE compiling 254.gap on ia64 with -O2 -funroll-loops
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          ice-on-legal-code
>Submitter-Id:   net
>Arrival-Date:   Thu Mar 14 14:26:01 PST 2002
>Closed-Date:
>Last-Modified:
>Originator:     Janis Johnson
>Release:        gcc version 3.1 20020314 (prerelease)
>Organization:
>Environment:
Itanium, Red Hat Linux release 7.1.94 (Roswell)
>Description:
GCC 3.1 prerelease gets an ICE compiling 254.gap from SPEC
CPU2000 with "-O2 -funroll-loops" on ia64-unknown-linux-gnu.
Compiler output is:

bug.c: In function `f':
bug.c:49: Attempt to delete prologue/epilogue insn:
(insn 553 551 555 (set (reg:DI 333 ar.lc)
        (reg:DI 34 r35)) -1 (nil)
    (nil))
bug.c:49: Internal compiler error in propagate_one_insn, at flow.c:1573
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.

This is a regression from GCC 3.0.4.
>How-To-Repeat:
Compile the test case on ia64 with -O2 -funroll-loops.
>Fix:
Unknown
>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: text/plain; name="bug.c"
Content-Disposition: inline; filename="bug.c"

/* This test case is extracted from file costab.c in 254.gap in SPEC
   CPU2000.  It's been cut down significantly and the names have been
   changed.  The ICE seen when compiling the original file is the same
   with this version.  That ICE occurs with gcc version 3.1 20020314
   (prerelease) on ia64-unknown-linux-gnu.

   From the original file: */

/***************************************************************************
**
*A  costab.c                    GAP source                   Martin Schoenert
*A                                                           & Volkmar Felsch
**
*Y  Copyright 1990-1992,  Lehrstuhl D fuer Mathematik,  RWTH Aachen,  Germany
**
****************************************************************************
*/

struct S {int i;};
typedef struct S * T;

extern T Error (const char *);
static long lenw;

T
f (T hdCall)
{
  T * p1;
  T * p2;
  T * p3;
  long n = 10;
  long leng;
  long i, j, k = 3;
  long gen = 0;

  if ((long)p3[0] != lenw)
    return Error( "message 1");

  if ((long)p3[k] < 0)
    for ( i = k; i <= leng; i++)
      p3[i] = (T)(-(long)p3[i]);

  for (i = 1; i <= n; i++)
    if ((long)p1[i] <= -i || (long)p1[i] >= i
        || (long)p2[i] <= -i || (long)p2[i] >= i )
      return Error ("message 2");

  return (T)gen;
}