[Bug target/13918] New: [3.3.3 regression] [ia64] bootstrap comparision failure

[Bug target/13918] New: [3.3.3 regression] [ia64] bootstrap comparision failure

[debian-gcc at lists dot debian dot org]

(environment is Debian unstable, binutils 2.14.90.0.7)

3.3.3 CVS 20040128 fails in the boostrap comparision (java/decl.o), CVS 20040110
 doesn't show the failure.

--- decl-stage2.s       2004-01-29 06:28:48.000000000 +0000
+++ decl-stage3.s       2004-01-29 06:28:58.000000000 +0000
@@ -1,5 +1,5 @@
 
-stage2/java/decl.o:     file format elf64-ia64-little
+java/decl.o:     file format elf64-ia64-little
 
 Disassembly of section .text:
 
@@ -2051,10 +2051,10 @@
     2a40:      0b 70 00 1e 18 10       [MMI]       ld8 r14=[r15];;
     2a46:      70 00 38 0c 72 00                   cmp.eq p7,p6=0,r14
     2a4c:      00 00 04 00                         nop.i 0x0;;
-    2a50:      dc 00 20 40 98 11       [MFB] (p06) st8 [r32]=r8
+    2a50:      fc 00 20 1e 98 11       [MFB] (p07) st8 [r15]=r8
     2a56:      00 00 00 02 00 00                   nop.f 0x0
     2a5c:      00 00 00 20                         nop.b 0x0
-    2a60:      e0 00 20 1e 98 11       [MII] (p07) st8 [r15]=r8
+    2a60:      c0 00 20 40 98 11       [MII] (p06) st8 [r32]=r8
     2a66:      f0 c0 22 00 42 00                   adds r15=88,r8
     2a6c:      04 40 00 84                         mov r32=r8
     2a70:      0a 70 00 46 18 10       [MMI]       ld8 r14=[r35];;

-- 
           Summary: [3.3.3 regression] [ia64] bootstrap comparision failure
           Product: gcc
           Version: 3.3.3
            Status: UNCONFIRMED
          Severity: critical
          Priority: P2
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: debian-gcc at lists dot debian dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: ia64-linux
  GCC host triplet: ia64-linux
GCC target triplet: ia64-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-29 06:41 -------
Confirmed, I think there is a known bug some where in the optimizers which cause this, it only 
effects ia64 because it abuses one of the passes cannot remember which one.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2004-01-29 06:41:25
               date|                            |
   Target Milestone|---                         |3.3.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-29 06:49 -------
See http://gcc.gnu.org/ml/gcc/2003-07/msg00585.html.  This is caused by IA64 abuse of the 
scheduler.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-29 06:55 -------
Based a discusion with Zack, this is minor as yyou can still continue from this with just make and 
the compiler will still work right as it is just two instructions have been interchanged and in this 
case the order does not matter.  see <http://gcc.gnu.org/ml/gcc/2003-07/msg00780.html> for 
even more info about this bug. (do not trust H. J. Lu on his analysis of the problem).

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |minor


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-29 06:59 -------
Could you try with this patch: <http://gcc.gnu.org/ml/gcc-patches/2004-01/msg02931.html>, 
the problem is in cselib which gets abused by ia64.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[wilson at gcc dot gnu dot org]

------- Additional Comments From wilson at gcc dot gnu dot org  2004-01-29 21:41 -------
This is a known interaction problem with cselib/sched-ebb.  I have never seen a
report of it resulting in an ICE or incorrect code.  It only results in code
that is scheduled differently, and hence this can cause the gcc compare test to
fail.

The oversimplified description of the problem is that memory is being accessed
after it has been freed.

This problem is difficult to reproduce.  It is known to be sensitive to the
environment size.  Small changes to the environment size can cause the problem
to appear or disappear.  This is because environment size changes cause pointers
to have different values resulting in different behaviour in the presense of the
bug.

The problem is difficult to fix.  Andrew Pinksi suggested a patch, but it is one
patch is a long sequence of patches that have been directed at this problem, and
the problem is still not fully fixed on mainline.  I do not recommend taking
this approach for gcc-3.3.

It is safe to ignore the problem if you can.  If you can't ignore the problem
then I would suggest changing sched-ebb.c to stop using cselib.  This will
result in some performance loss, though I don't know how much.  There is a minor
chance of exposing latent bugs with this change, but I think the chance is
acceptably small.

I will put my suggested patch in an attachment.  If this patch does not fix the
problem, then there must be something else wrong that needs to be investigated.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[gdr at gcc dot gnu dot org]

------- Additional Comments From gdr at gcc dot gnu dot org  2004-02-15 12:43 -------
Adjust milestone

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|3.3.3                       |3.3.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[schwab at suse dot de]

------- Additional Comments From schwab at suse dot de  2004-02-22 16:57 -------
FWIW, I'm using this patch in the daily regression test and haven't seen a 
comparison failure since then. 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2004-02-22 17:01 -------
Subject: Re:  [3.3.3 regression] [ia64] bootstrap comparision failure

"schwab at suse dot de" <gcc-bugzilla@gcc.gnu.org> writes:

| FWIW, I'm using this patch in the daily regression test and haven't seen a 
| comparison failure since then. 

That is good to know.  Would mind going ahead and apply it to
gcc-3_3-branch? Thanks,

-- Gaby
 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-02-22 17:56 -------
Subject: Bug 13918

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_3-branch
Changes by:	schwab@gcc.gnu.org	2004-02-22 17:56:50

Modified files:
	gcc            : ChangeLog sched-ebb.c 

Log message:
	2004-02-22  James E Wilson  <wilson@specifixinc.com>
	
	PR target/13918
	* sched-ebb.c (ebb_sched_info): Disable use of cselib.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.16114.2.903&r2=1.16114.2.904
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/sched-ebb.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.17.20.1&r2=1.17.20.2



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918

[Bug target/13918] [3.3.3 regression] [ia64] bootstrap comparision failure

[schwab at suse dot de]

------- Additional Comments From schwab at suse dot de  2004-02-22 17:59 -------
Checked in. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13918