[Bug rtl-optimization/52250] New: [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[Bug rtl-optimization/52250] New: [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[zsojka at seznam dot cz]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

             Bug #: 52250
           Summary: [4.7 Regression] ICE: in sel_remove_bb, at
                    sel-sched-ir.c:5213 with -fsel-sched-pipelining
                    -fsel-sched-pipelining-outer-loops
                    -fselective-scheduling2 and other flags
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: zsojka@seznam.cz
              Host: x86_64-pc-linux-gnu
            Target: x86_64-pc-linux-gnu


Created attachment 26661
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26661
reduced testcase

Once again an ICE with insane compiler flags.

Compiler output:
$ gcc -O -ftree-pre -fno-guess-branch-probability -freorder-blocks
-fschedule-insns2 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops
-fselective-scheduling2 -fno-tree-dominator-opts -fno-tree-forwprop testcase.f
testcase.f: In function 'foo':
testcase.f:12:0: internal compiler error: in sel_remove_bb, at
sel-sched-ir.c:5213
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

(gdb) bt
#0  fancy_abort (file=0x130f4a0 "/mnt/svn/gcc-trunk/gcc/sel-sched-ir.c",
line=5213, function=0x13100a0 "sel_remove_bb")
    at /mnt/svn/gcc-trunk/gcc/diagnostic.c:898
#1  0x0000000000a06c41 in sel_remove_bb (bb=0x7ffff5ac6680,
remove_from_cfg_p=<optimized out>) at
/mnt/svn/gcc-trunk/gcc/sel-sched-ir.c:5213
#2  0x0000000000a0f237 in maybe_tidy_empty_bb (bb=0x7ffff5ac6680) at
/mnt/svn/gcc-trunk/gcc/sel-sched-ir.c:3753
#3  0x0000000000a0f639 in tidy_control_flow (xbb=0x7ffff5ac6680,
full_tidying=true) at /mnt/svn/gcc-trunk/gcc/sel-sched-ir.c:3776
#4  0x0000000000a0ff46 in sel_remove_insn (insn=0x7ffff5bc8d38,
only_disconnect=false, full_tidying=true)
    at /mnt/svn/gcc-trunk/gcc/sel-sched-ir.c:3923
#5  0x0000000000a1040a in return_nop_to_pool (nop=0x7ffff5bc8d38,
full_tidying=<optimized out>) at /mnt/svn/gcc-trunk/gcc/sel-sched-ir.c:1061
#6  0x0000000000a254ba in remove_temp_moveop_nops (full_tidying=true) at
/mnt/svn/gcc-trunk/gcc/sel-sched.c:4991
#7  schedule_expr_on_boundary (bnd=0x1ad0fb0, expr_vliw=0x1ad1910, seqno=-1) at
/mnt/svn/gcc-trunk/gcc/sel-sched.c:5467
#8  0x0000000000a26eb4 in fill_insns (scheduled_insns_tailpp=<synthetic
pointer>, seqno=-1, fence=<optimized out>)
    at /mnt/svn/gcc-trunk/gcc/sel-sched.c:5584
#9  schedule_on_fences (scheduled_insns_tailpp=<synthetic pointer>,
max_seqno=9, fences=0x1ad0cd8) at /mnt/svn/gcc-trunk/gcc/sel-sched.c:7343
#10 sel_sched_region_2 (orig_max_seqno=21) at
/mnt/svn/gcc-trunk/gcc/sel-sched.c:7477
#11 0x0000000000a2aa55 in sel_sched_region_1 () at
/mnt/svn/gcc-trunk/gcc/sel-sched.c:7516
#12 sel_sched_region (rgn=1) at /mnt/svn/gcc-trunk/gcc/sel-sched.c:7617
#13 0x0000000000a2c2ba in run_selective_scheduling () at
/mnt/svn/gcc-trunk/gcc/sel-sched.c:7693
#14 0x00000000009f90f6 in rest_of_handle_sched2 () at
/mnt/svn/gcc-trunk/gcc/sched-rgn.c:3515
#15 0x0000000000968d75 in execute_one_pass (pass=0x1852b00) at
/mnt/svn/gcc-trunk/gcc/passes.c:2081
#16 0x0000000000969135 in execute_pass_list (pass=0x1852b00) at
/mnt/svn/gcc-trunk/gcc/passes.c:2136
#17 0x0000000000969147 in execute_pass_list (pass=0x1852100) at
/mnt/svn/gcc-trunk/gcc/passes.c:2137
#18 0x0000000000969147 in execute_pass_list (pass=0x1852160) at
/mnt/svn/gcc-trunk/gcc/passes.c:2137
#19 0x0000000000acb4ee in tree_rest_of_compilation (fndecl=0x7ffff5bb5300) at
/mnt/svn/gcc-trunk/gcc/tree-optimize.c:422
#20 0x000000000071da4a in cgraph_expand_function (node=0x7ffff5a95900) at
/mnt/svn/gcc-trunk/gcc/cgraphunit.c:1831
#21 0x000000000071f93c in cgraph_expand_all_functions () at
/mnt/svn/gcc-trunk/gcc/cgraphunit.c:1898
#22 cgraph_optimize () at /mnt/svn/gcc-trunk/gcc/cgraphunit.c:2212
#23 0x000000000071ff8a in cgraph_finalize_compilation_unit () at
/mnt/svn/gcc-trunk/gcc/cgraphunit.c:1338
#24 0x0000000000912fed in write_global_declarations () at
/mnt/svn/gcc-trunk/gcc/langhooks.c:303
#25 0x0000000000a5dc1c in compile_file () at
/mnt/svn/gcc-trunk/gcc/toplev.c:573
#26 do_compile () at /mnt/svn/gcc-trunk/gcc/toplev.c:1938
#27 toplev_main (argc=24, argv=0x7fffffffd628) at
/mnt/svn/gcc-trunk/gcc/toplev.c:2014
#28 0x00007ffff61cc09d in __libc_start_main () from /lib64/libc.so.6
#29 0x0000000000584ea1 in _start ()

Tested revisions:
r184192 - crash
r180540 - crash
r178498 - OK
4.6 r180325 - OK

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-02-20
     Ever Confirmed|0                           |1

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> 2012-02-20 22:25:25 UTC ---
Confirmed.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-02-22 16:45:03 UTC ---
Started with http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=179284
which doesn't seem to be related, so looks like a latent sel-sched bug.
The note the assert fails on is NOTE_INSN_DELETED, so one has to wonder why we
don't throw such notes away during the scheduling, or at least don't fail the
assertion on those.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

--- Comment #3 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-02-22 18:49:15 UTC ---
I'll take a look on Friday, tomorrow I am off work. Indeed I do not remember
seeing these notes at the assert time.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

--- Comment #4 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-27 10:39:32 UTC ---
Surely not P1 at this point.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

--- Comment #5 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-02-27 13:03:58 UTC ---
When removing an empty block, we need to find its neighbour to stick the bb
note list to, and the code doing this now just picks some pred block from the
loop that redirects preds' edges around the block being removed.  Which happens
to be the wrong one for the test case.  Fixed by making the logic properly
consider preds and the only succ, if needed, as below.  I will test this
tomorrow and still can make it before the RC1, I think.

diff --git a/gcc/sel-sched-ir.c b/gcc/sel-sched-ir.c
index a93cd68..03d135f 100644
--- a/gcc/sel-sched-ir.c
+++ b/gcc/sel-sched-ir.c
@@ -3658,7 +3658,7 @@ sel_recompute_toporder (void)
 static bool
 maybe_tidy_empty_bb (basic_block bb)
 {
-  basic_block succ_bb, pred_bb;
+  basic_block succ_bb, pred_bb, note_bb;
   VEC (basic_block, heap) *dom_bbs;
   edge e;
   edge_iterator ei;
@@ -3697,6 +3697,17 @@ maybe_tidy_empty_bb (basic_block bb)
   pred_bb = NULL;
   dom_bbs = NULL;

+  /* Save a pred/succ from the current region to attach the notes to.  */
+  note_bb = NULL;
+  FOR_EACH_EDGE (e, ei, bb->preds)
+    if (in_current_region_p (e->src))
+      {
+    note_bb = e->src;
+    break;
+      }
+  if (note_bb == NULL)
+    note_bb = succ_bb;
+
   /* Redirect all non-fallthru edges to the next bb.  */
   while (rescan_p)
     {
@@ -3746,10 +3757,8 @@ maybe_tidy_empty_bb (basic_block bb)
   else
     {
       /* This is a block without fallthru predecessor.  Just delete it.  */
-      gcc_assert (pred_bb != NULL);
-
-      if (in_current_region_p (pred_bb))
-    move_bb_info (pred_bb, bb);
+      gcc_assert (note_bb);
+      move_bb_info (note_bb, bb);
       remove_empty_bb (bb, true);
     }

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Andrey Belevantsev <abel at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
         AssignedTo|unassigned at gcc dot       |abel at gcc dot gnu.org
                   |gnu.org                     |

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[steven at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Steven Bosscher <steven at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |steven at gcc dot gnu.org

--- Comment #6 from Steven Bosscher <steven at gcc dot gnu.org> 2012-02-27 22:16:02 UTC ---
Out of curiosity: Is it possible to bootstrap x86_64-*-linux-gnu with the
selective scheduler enabled (instead of sched2), and if so, has anyone ever
benchmarked the effect?
(I know, not on topic in this PR, but...)

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Andrey Belevantsev <abel at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amonakov at gcc dot gnu.org

--- Comment #7 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-02-28 09:04:46 UTC ---
It should be possible to bootstrap, we generally keep an eye on it.  Last time
we did benchmarking I think was when Alexander implemented predication support,
as it works on x86-64 as well (and helps pipelining there, too).  Probably the
slight improvement wasn't worth the slowdown, but Alexander should remember
better for sure.  And of course this can be re-benchmarked on Stage 1 with
predication committed and enabled.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

--- Comment #8 from Alexander Monakov <amonakov at gcc dot gnu.org> 2012-02-28 10:20:58 UTC ---
Like Andrey said, we verify that x86_64-linux bootstraps with sel-sched when
submitting patches, but I do not remember any specific figures from the times
when I did benchmarking.  It's a good suggestion do to it again in the next
stage 1.

[Bug rtl-optimization/52250] [4.7/4.8 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

--- Comment #9 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-03-06 07:31:49 UTC ---
Author: abel
Date: Tue Mar  6 07:31:43 2012
New Revision: 184975

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=184975
Log:
        PR rtl-optimization/52250
        * sel-sched-ir.c (maybe_tidy_empty_bb): Try harder to find a bb
        to put note list into.  Unconditionally call move_bb_info.
        (move_bb_info): Do not assert the blocks being in the same region,
        just drop the note list if they are not.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/sel-sched-ir.c

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Andrey Belevantsev <abel at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|                            |4.8.0
   Target Milestone|4.7.0                       |4.7.1
            Summary|[4.7/4.8 Regression] ICE:   |[4.7 Regression] ICE: in
                   |in sel_remove_bb, at        |sel_remove_bb, at
                   |sel-sched-ir.c:5213 with    |sel-sched-ir.c:5213 with
                   |-fsel-sched-pipelining      |-fsel-sched-pipelining
                   |-fsel-sched-pipelining-oute |-fsel-sched-pipelining-oute
                   |r-loops                     |r-loops
                   |-fselective-scheduling2 and |-fselective-scheduling2 and
                   |other flags                 |other flags

--- Comment #10 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-03-06 08:02:27 UTC ---
Fixed on trunk, the 4.7 branch will be fixed after 4.7.0.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.7.1                       |4.7.2

--- Comment #11 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-06-14 08:49:35 UTC ---
GCC 4.7.1 is being released, adjusting target milestone.

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[amonakov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aturjan at yahoo dot com

--- Comment #12 from Alexander Monakov <amonakov at gcc dot gnu.org> 2012-07-17 16:08:13 UTC ---
*** Bug 53999 has been marked as a duplicate of this bug. ***

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

--- Comment #13 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-07-20 08:49:54 UTC ---
Author: abel
Date: Fri Jul 20 08:49:52 2012
New Revision: 189705

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=189705
Log:
        Backport from mainline
        2012-03-06  Andrey Belevantsev  <abel@ispras.ru>

        PR rtl-optimization/52250
        * sel-sched-ir.c (maybe_tidy_empty_bb): Try harder to find a bb
        to put note list into.  Unconditionally call move_bb_info.
        (move_bb_info): Do not assert the blocks being in the same region,
        just drop the note list if they are not.


Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/sel-sched-ir.c

[Bug rtl-optimization/52250] [4.7 Regression] ICE: in sel_remove_bb, at sel-sched-ir.c:5213 with -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops -fselective-scheduling2 and other flags

[abel at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52250

Andrey Belevantsev <abel at gcc dot gnu.org> changed:

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

--- Comment #14 from Andrey Belevantsev <abel at gcc dot gnu.org> 2012-07-20 08:52:39 UTC ---
The duplicate PR reminded me of the backporting, so fixed on 4.7 branch as well
and closed.