[Bug rtl-optimization/60317] New: [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[Bug rtl-optimization/60317] New: [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[trippels at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

            Bug ID: 60317
           Summary: [4.9 Regression] find_hard_regno_for compile time hog
                    in libvpx
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: trippels at gcc dot gnu.org
                CC: vmakarov at gcc dot gnu.org

While building Firefox I've noticed that there is a file from libvpx
that takes almost 2 minutes to compile on my machine:

 % time gcc -O3 -c sad_c.i
gcc -O3 -c sad_c.i  111.05s user 0.78s system 99% cpu 1:52.10 total

 phase opt and generate  : 110.92 (100%) usr   0.43 (100%) sys 111.43 (100%)
wall  168173 kB (99%) ggc

Perf shows:
 87.86%  cc1  cc1   [.] find_hard_regno_for(int, int*, int)
   - find_hard_regno_for(int, int*, int)
        99.77% 0x10101010101
  0.72%  cc1  cc1   [.] find_base_term(rtx_def*)


It's fine with -O2 or 4.8.3:

 % time gcc -O2 -c sad_c.i
gcc -O2 -c sad_c.i  0.25s user 0.02s system 99% cpu 0.273 total

 phase opt and generate  :   0.23 (96%) usr   0.02 (100%) sys   0.25 (93%) wall
   8876 kB (84%) ggc

 % time /usr/x86_64-pc-linux-gnu/gcc-bin/4.8.3/gcc -w -O3 -c sad_c.i
/usr/x86_64-pc-linux-gnu/gcc-bin/4.8.3/gcc -w -O3 -c sad_c.i  0.94s user 0.02s
system 99% cpu 0.964 total

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[trippels at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

--- Comment #1 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Created attachment 32200
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32200&action=edit
preprocessed testcase

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[trippels at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |compile-time-hog
   Last reconfirmed|                            |2014-2-22
      Known to work|                            |4.7.3, 4.8.3
   Target Milestone|---                         |4.9.0
      Known to fail|                            |4.9.0

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[trippels at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|2014-02-22 00:00:00         |2014-02-23
                 CC|                            |congh at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Started with r204229.

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
This whole thing updating keys and such should use a proper lattice of
per-cgraph and per-edge node sizes/times which can be updated with a
less ad-hoc algorithm than the current one which easily and completely explodes
... :/  (and the 'cache' is completely ineffective during the update process)

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #3)
> This whole thing updating keys and such should use a proper lattice of
> per-cgraph and per-edge node sizes/times which can be updated with a
> less ad-hoc algorithm than the current one which easily and completely
> explodes ... :/  (and the 'cache' is completely ineffective during the
> update process)

Err - wrong bug

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[vmakarov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

--- Comment #5 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
Author: vmakarov
Date: Tue Feb 25 20:34:44 2014
New Revision: 208155

URL: http://gcc.gnu.org/viewcvs?rev=208155&root=gcc&view=rev
Log:
2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>

    PR rtl-optimization/60317
    * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
    * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
    * lra-assigns.c: Include params.h.
    (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
    other reload pseudos considerations.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/lra-assigns.c
    trunk/gcc/params.def
    trunk/gcc/params.h

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[trippels at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317

Markus Trippelsdorf <trippels at gcc dot gnu.org> changed:

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

--- Comment #6 from Markus Trippelsdorf <trippels at gcc dot gnu.org> ---
Fixed, thanks.

 % time gcc -c -O3 sad_c.i
gcc -c -O3 sad_c.i  12.72s user 0.11s system 99% cpu 12.837 total

(Vladimir, I don't see any message on gcc-patches about your fix.)

[Bug rtl-optimization/60317] [4.9 Regression] find_hard_regno_for compile time hog in libvpx

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60317
Bug 60317 depends on bug 60343, which changed state.

Bug 60343 Summary: [4.9 Regression] r208155 breaks bootstrap
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60343

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