[Bug rtl-optimization/58826] New: Runfail on CPU2006 436.cactusADM with after r203739 for core-avx2 target.

[Bug rtl-optimization/58826] New: Runfail on CPU2006 436.cactusADM with after r203739 for core-avx2 target.

[ysrumyan at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

            Bug ID: 58826
           Summary: Runfail on CPU2006 436.cactusADM with  after r203739
                    for core-avx2 target.
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ysrumyan at gmail dot com

To reproduce it is sufficient to compile only one source file -
StaggeredLeapfrog2.F which must be preprocessed. In rtl-dump after reload we
can see the following erroneous of instruction which try to write to RO
section:

(insn 10377 15507 10379 49 (set (mem/u/c:V4DF (symbol_ref/u:DI ("*.LC10")
[flags 0x2]) [3 S32 A256])
        (reg:V4DF 27 xmm6 [10848])) 1158 {*movv4df_internal}
     (nil))

The following options were used to compile:

-march=core-avx2 -Ofast  -funroll-loops -fno-second-underscore

although the last option looks redundant.

[Bug rtl-optimization/58826] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2013-10-21
           Assignee|unassigned at gcc dot gnu.org      |rguenth at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I will have a look though I doubt I caused this (if you bisect past that
revision
and revert the PR58464 fix as well)

[Bug rtl-optimization/58826] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[ysrumyan at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

--- Comment #2 from Yuri Rumyantsev <ysrumyan at gmail dot com> ---
In fact LRA is responsible for this failure - there is a bug in constant
regeneration. LRA correctly regenerates all occurrences of virtual register
which is not allocated(i.e. it does not has a register) but it must delete its
spill/reload since this virtual register is dead.

[Bug rtl-optimization/58826] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ra
             Status|ASSIGNED                    |NEW
                 CC|                            |vmakarov at gcc dot gnu.org
           Assignee|rguenth at gcc dot gnu.org         |unassigned at gcc dot gnu.org

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Not mine then.

[Bug rtl-optimization/58826] [4.9 Regression] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
   Target Milestone|---                         |4.9.0
            Summary|Runfail on CPU2006          |[4.9 Regression] Runfail on
                   |436.cactusADM with  after   |CPU2006 436.cactusADM with
                   |r203377 for core-avx2       |after r203377 for core-avx2
                   |target.                     |target.

[Bug rtl-optimization/58826] [4.9 Regression] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[vmakarov at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

--- Comment #4 from Vladimir Makarov <vmakarov at gcc dot gnu.org> ---
(In reply to Yuri Rumyantsev from comment #2)
> In fact LRA is responsible for this failure - there is a bug in constant
> regeneration. LRA correctly regenerates all occurrences of virtual register
> which is not allocated(i.e. it does not has a register) but it must delete
> its spill/reload since this virtual register is dead.

Could you check a fix for PR58759 I included into the trunk recently.

Thanks.

[Bug rtl-optimization/58826] [4.9 Regression] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Waiting for reporter.

[Bug rtl-optimization/58826] [4.9 Regression] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[ysrumyan at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

--- Comment #6 from Yuri Rumyantsev <ysrumyan at gmail dot com> ---
I assume that this bug should be closed since it is not reproducible after the
latest LRA fix.

[Bug rtl-optimization/58826] [4.9 Regression] Runfail on CPU2006 436.cactusADM with after r203377 for core-avx2 target.

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58826

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

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

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed thus.