Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[zack]

Old Synopsis: [3.3 regression] [HP-PA] inline optimization ICE
New Synopsis: [3.4 regression] [HP-PA] inline optimization ICE

Responsible-Changed-From-To: unassigned->hubicka
Responsible-Changed-By: zack
Responsible-Changed-When: Sun Apr  6 00:47:52 2003
Responsible-Changed-Why:
    Now fixed in 3.3; Jan is working on a 3.4 fix.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10024

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[John David Anglin]

The following reply was made to PR optimization/10024; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: jh@suse.cz (Jan Hubicka)
Cc: zack@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org,
   tausq@debian.org
Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
Date: Fri, 11 Apr 2003 11:21:08 -0400 (EDT)

 > > It appears that the addition of the following patch
 > > 
 > > Mon Apr  7 15:56:30 CEST 2003  Jan Hubicka  <jh@suse.cz>
 > > 
 > >         PR opt/10024
 > > 	Revert Zack's change
 > > 	* cfglayout.c (cfg_layout_redirect_edge):
 > > 	Redirect any branch edges unified with the fallthru one.
 > > 	* cfgrtl.c (force_nonfallthru_and_redirect):  Do not special
 > > 	case fallthru edges when called via cfglayout.c
 > > 
 > > causes gcc.c-torture/compile/20030405-1.c to fail on
 > > hppa2.0w-hp-hpux11.11 on the 3.3 branch at -O2 and above.
 
 The failure only exists on 3.3, so I suspect the fix must already
 be in the trunk.
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[hubicka]

Synopsis: [3.4 regression] [HP-PA] inline optimization ICE

State-Changed-From-To: analyzed->closed
State-Changed-By: hubicka
State-Changed-When: Mon Apr 14 17:27:00 2003
State-Changed-Why:
    Commit fix as obvious thus.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10024

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[John David Anglin]

The following reply was made to PR optimization/10024; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: jh@suse.cz (Jan Hubicka)
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
Date: Sat, 12 Apr 2003 23:47:16 -0400 (EDT)

 Hi Jan,
 
 The patch fixes the PR10024 testsuite fails.  I have tested it with full
 builds on hppa2.0w-hp-hpux11.11 3.4; hppa2.0w-hp-hpux11.11, hppa64-hp-hpux11.11
 and hppa-unknown-linux-gnu 3.3.  There were no regressions in any of the tests.
 
 Would you please submit it for review?
 
 Thanks.
 
 > > The following reply was made to PR optimization/10024; it has been noted by GNATS.
 > > 
 > > From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
 > > To: jh@suse.cz (Jan Hubicka)
 > > Cc: jh@suse.cz, zack@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org,
 > >    tausq@debian.org
 > > Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
 > > Date: Fri, 11 Apr 2003 13:41:48 -0400 (EDT)
 > > 
 > >  > Actually it is not.  Only trunk has different reordering alrotihm.
 > >  > I've never heard back about the bootstrap sucess/failure with my
 > >  > proposed patch.  Will try to check the archives.
 > >  
 > >  I haven't seen a patch.  Could you resend?
 > Hope this is the right one.
 > 
 > Index: cfgrtl.c
 > ===================================================================
 > RCS file: /cvs/gcc/gcc/gcc/cfgrtl.c,v
 > retrieving revision 1.61.2.6
 > diff -c -3 -p -r1.61.2.6 cfgrtl.c
 > *** cfgrtl.c	7 Apr 2003 16:31:42 -0000	1.61.2.6
 > --- cfgrtl.c	11 Apr 2003 21:34:02 -0000
 > *************** force_nonfallthru_and_redirect (e, targe
 > *** 949,955 ****
 >         && JUMP_LABEL (e->src->end) == e->dest->head)
 >       {
 >         rtx note;
 > !       edge b = make_edge (e->src, target, 0);
 >   
 >         if (!redirect_jump (e->src->end, block_label (target), 0))
 >   	abort ();
 > --- 949,955 ----
 >         && JUMP_LABEL (e->src->end) == e->dest->head)
 >       {
 >         rtx note;
 > !       edge b = unchecked_make_edge (e->src, target, 0);
 >   
 >         if (!redirect_jump (e->src->end, block_label (target), 0))
 >   	abort ();
 > 
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[John David Anglin]

The following reply was made to PR optimization/10024; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: jh@suse.cz (Jan Hubicka)
Cc: jh@suse.cz, zack@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org,
   tausq@debian.org
Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
Date: Fri, 11 Apr 2003 13:41:48 -0400 (EDT)

 > Actually it is not.  Only trunk has different reordering alrotihm.
 > I've never heard back about the bootstrap sucess/failure with my
 > proposed patch.  Will try to check the archives.
 
 I haven't seen a patch.  Could you resend?
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[Jan Hubicka]

The following reply was made to PR optimization/10024; it has been noted by GNATS.

From: Jan Hubicka <jh@suse.cz>
To: John David Anglin <dave@hiauly1.hia.nrc.ca>
Cc: Jan Hubicka <jh@suse.cz>, zack@gcc.gnu.org, gcc-bugs@gcc.gnu.org,
	gcc-gnats@gcc.gnu.org, tausq@debian.org
Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
Date: Fri, 11 Apr 2003 18:58:20 +0200

 > > > It appears that the addition of the following patch
 > > > 
 > > > Mon Apr  7 15:56:30 CEST 2003  Jan Hubicka  <jh@suse.cz>
 > > > 
 > > >         PR opt/10024
 > > > 	Revert Zack's change
 > > > 	* cfglayout.c (cfg_layout_redirect_edge):
 > > > 	Redirect any branch edges unified with the fallthru one.
 > > > 	* cfgrtl.c (force_nonfallthru_and_redirect):  Do not special
 > > > 	case fallthru edges when called via cfglayout.c
 > > > 
 > > > causes gcc.c-torture/compile/20030405-1.c to fail on
 > > > hppa2.0w-hp-hpux11.11 on the 3.3 branch at -O2 and above.
 > 
 > The failure only exists on 3.3, so I suspect the fix must already
 > be in the trunk.
 
 Actually it is not.  Only trunk has different reordering alrotihm.
 I've never heard back about the bootstrap sucess/failure with my
 proposed patch.  Will try to check the archives.
 
 Honza
 > 
 > Dave
 > -- 
 > J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
 > National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)

Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE

[John David Anglin]

The following reply was made to PR optimization/10024; it has been noted by GNATS.

From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
To: zack@gcc.gnu.org, gcc-bugs@gcc.gnu.org, gcc-gnats@gcc.gnu.org, jh@suse.cz,
   tausq@debian.org
Cc:  
Subject: Re: optimization/10024: [3.4 regression] [HP-PA] inline optimization ICE
Date: Mon, 7 Apr 2003 19:18:51 -0400 (EDT)

 It appears that the addition of the following patch
 
 Mon Apr  7 15:56:30 CEST 2003  Jan Hubicka  <jh@suse.cz>
 
         PR opt/10024
 	Revert Zack's change
 	* cfglayout.c (cfg_layout_redirect_edge):
 	Redirect any branch edges unified with the fallthru one.
 	* cfgrtl.c (force_nonfallthru_and_redirect):  Do not special
 	case fallthru edges when called via cfglayout.c
 
 causes gcc.c-torture/compile/20030405-1.c to fail on
 hppa2.0w-hp-hpux11.11 on the 3.3 branch at -O2 and above.
 
 <http://gcc.gnu.org/ml/gcc-testresults/2003-04/msg00438.html>.
 
 The test passed in my previous build.
 
 Running under gdb, the crash occurs at the spot mentioned by Zack
 in the audit trail:
 
 Starting program: /xxx/gnu/gcc-3.3/objdir/gcc/testsuite/../cc1 -O2 /xxx/gnu/gcc-3.3/gcc/gcc/testsuite/gcc.c-torture/compile/20030405-1.c
 warning: Unable to find __d_pid symbol in object file.
 warning: Suggest linking with /opt/langtools/lib/end.o.
 warning: GDB will be unable to track shl_load/shl_unload calls
  fixfloor fixtoi ftofix fixtof fixdiv itofix foo
 Program received signal SIGSEGV, Segmentation fault.
 force_nonfallthru_and_redirect (e=0x4008a000, target=0x40089a60)
     at ../../gcc/gcc/cfgrtl.c:961
 961               b->probability = prob;
 (gdb) p b
 $1 = 0x0
 
 Dave
 -- 
 J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
 National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)