[Bug tree-optimization/22471] New: corrupted profile info with -O3 -fprofile-use

[Bug tree-optimization/22471] New: corrupted profile info with -O3 -fprofile-use

[janis at gcc dot gnu dot org]

Several of the SPEC CPU2000 tests currently fail on powerpc64-linux
with -fprofile-generate/-fprofile-use plus -O3.  Some of the failures
start recently and appear to be intermittent.  Failures for 176.gcc
with either -m32 or -m64 and 175.vpr with -m64 start with this patch
from dberlin (Remove var_ann->uid) and seem to fail consistently since
it was added:
                                                                                
  http://gcc.gnu.org/ml/gcc-cvs/2005-06/msg01108.html
                                                                                
Error messages are "corrupted profile info: number of executions for
edge xxx-yyy thought to be zzz" or "number of iterations for basic
block xxx thought to be yyy" or "coverage mismatch for function 'xxx'
while reading counter 'arcs'".
                                                                                
I don't have a minimized test case but can try to create one.

-- 
           Summary: corrupted profile info with -O3 -fprofile-use
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: tree-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: janis at gcc dot gnu dot org
                CC: dberlin at gcc dot gnu dot org,gcc-bugs at gcc dot gnu
                    dot org
GCC target triplet: powerpc64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org
           Keywords|                            |ice-on-valid-code
            Summary|corrupted profile info with |[4.1 Regression] corrupted
                   |-O3 -fprofile-use           |profile info with -O3 -
                   |                            |fprofile-use
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[dberlin at dberlin dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-14 00:25 -------
Subject: Re:  New: corrupted profile info with
	-O3 -fprofile-use

On Wed, 2005-07-13 at 23:28 +0000, janis at gcc dot gnu dot org wrote:
> Several of the SPEC CPU2000 tests currently fail on powerpc64-linux
> with -fprofile-generate/-fprofile-use plus -O3.  Some of the failures
> start recently and appear to be intermittent.  Failures for 176.gcc
> with either -m32 or -m64 and 175.vpr with -m64 start with this patch
> from dberlin (Remove var_ann->uid) and seem to fail consistently since
> it was added:
>                                                                                 
>   http://gcc.gnu.org/ml/gcc-cvs/2005-06/msg01108.html

Are you certain it was this commit?
I can't for the life of me imagine why this patch would make any
difference.

Can you try the attached patch and see if it asserts while generating
profiling info?
This is the only  difference between the old way we did things and the
new way that i can think of that might matter, but i can't think why it
would only affect profiling code if so.



------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-14 00:25 -------
Created an attachment (id=9264)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9264&action=view)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[rguenth at gcc dot gnu dot org]

------- Additional Comments From rguenth at gcc dot gnu dot org  2005-07-14 11:06 -------
Doesn't trigger for the current profiledbootstrap problems (which look similar).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[janis at gcc dot gnu dot org]

------- Additional Comments From janis at gcc dot gnu dot org  2005-07-14 21:07 -------
The patch doesn't trigger anything, and I discovered that some of the failures
I'm seeing now happen before the patch from dberlin from 20050630.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[janis at gcc dot gnu dot org]

------- Additional Comments From janis at gcc dot gnu dot org  2005-07-15 00:07 -------
I've got diverging scripts on various systems, and when I thought I was testing
with a compiler from before 20050630 I was actually using today's mainline.

I have verified (and triple-checked) that none of the failures I currently see
in SPEC CPU2000 when using profile-generate/profile-use show up just before the
patch at http://gcc.gnu.org/ml/gcc-cvs/2005-06/msg01108.html.  Many of them show
up immediately after that patch, and others creep in later.  Among the failures
immediately after the patch are

  gap with profopts -O3 -m64
  gcc with profopts -O3 -m32/-m64
  twolf with profopts -O3 -m32/-m64
  vpr with profopts -O3 -m32/-m64

No minimized testcase yet.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[dberlin at dberlin dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-15 00:41 -------
Subject: Re:  [4.1 Regression] corrupted
	profile info with -O3 -fprofile-use

On Fri, 2005-07-15 at 00:07 +0000, janis at gcc dot gnu dot org wrote:
> ------- Additional Comments From janis at gcc dot gnu dot org  2005-07-15 00:07 -------
> I've got diverging scripts on various systems, and when I thought I was testing
> with a compiler from before 20050630 I was actually using today's mainline.
> 
> I have verified (and triple-checked) that none of the failures I currently see
> in SPEC CPU2000 when using profile-generate/profile-use show up just before the
> patch at http://gcc.gnu.org/ml/gcc-cvs/2005-06/msg01108.html.  Many of them show
> up immediately after that patch, and others creep in later.  Among the failures
> immediately after the patch are
> 
>   gap with profopts -O3 -m64
>   gcc with profopts -O3 -m32/-m64
>   twolf with profopts -O3 -m32/-m64
>   vpr with profopts -O3 -m32/-m64
> 
Okay, well.

I can eliminate 1/3rd of the patch's actual changes if you try the above
with -fno-tree-salias added.

Then, if that isn't better, can you also try the following patch and see
if it either triggers an assert or fixes the bug?

I never noticed before, but there are places we are looping over the
referenced variables and *adding things to the array at the same time*.
If the hashtable resizes at just the wrong point, i imagine this could
cause some bad things to happen.

> No minimized testcase yet.
> 


------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-15 00:41 -------
Created an attachment (id=9277)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9277&action=view)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 GCC target triplet|powerpc64-unknown-linux-gnu |powerpc64-*-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[steven at gcc dot gnu dot org]

------- Additional Comments From steven at gcc dot gnu dot org  2005-07-26 23:32 -------
Is sthis still an issue? 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[dberlin at dberlin dot org]

------- Additional Comments From dberlin at gcc dot gnu dot org  2005-07-26 23:45 -------
Subject: Re:  [4.1 Regression] corrupted
	profile info with -O3 -fprofile-use

Honza said he knew what was up and would fix it
On Tue, 2005-07-26 at 23:32 +0000, steven at gcc dot gnu dot org wrote:
> ------- Additional Comments From steven at gcc dot gnu dot org  2005-07-26 23:32 -------
> Is sthis still an issue? 
> 



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 14:15 -------
Is this fixed?  profiledbootstrap works for me.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[janis at gcc dot gnu dot org]

------- Additional Comments From janis at gcc dot gnu dot org  2005-08-11 17:29 -------
Mainline GCC built last night successfully compiles and runs the SPEC CPU2000
tests that failed earlier.  If anyone cares I can find out which patch fixed
it, otherwise we can just close this as magically fixed.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-11 17:31 -------
This was most likely fixed by the following patches:
2005-07-28  Jan Hubicka  <jh@suse.cz>

        * Makefile.in (rtl-profile.o): Kill all traces of it.
       ....
2005-08-03  Jan Hubicka  <jh@suse.cz>

        * cfgloop.h (DLTHE_FLAG_COMPLETTE_PEEL): New flag.
        * cfgloopmanip.c (duplicate_loop_to_header_edge): Special case
        profile updating for complette unrolling.
        * loop-unroll.c (peel_loop_completely): Use it.
        * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
        (tree_unroll_loops_completely): Disable code growing unrolling of cold
        loops.

2005-08-02  Jan Hubicka  <jh@suse.cz>

        * tree-ssa-dom.c (thread_across_edge): Remove updating here.
        * tree-ssa-threadupdate.c (thread_block): Add it here.

2005-08-01  Jan Hubicka  <jh@suse.cz>

        * profile.c (compute_value_histograms): Fix thinko.
        * value-prof.c: Include toplev.h
        (check_counter): New function.
        (tree_divmod_fixed_value_transform, tree_mod_pow2_value_transform,
        tree_mod_subtract_transform): Add sanity check.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471

[Bug tree-optimization/22471] [4.1 Regression] corrupted profile info with -O3 -fprofile-use

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-09-07 21:11 -------
Fixed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22471