[Bug target/66076] New: [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[Bug target/66076] New: [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[zsojka at seznam dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

            Bug ID: 66076
           Summary: [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618
                    with -funroll-loops -mno-prefer-avx128 -march=bdver4
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zsojka at seznam dot cz
  Target Milestone: ---

Created attachment 35501
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=35501&action=edit
reduced testcase

Compiler output:
$ gcc -O2 -ftree-loop-vectorize -funroll-loops -mno-prefer-avx128 -march=bdver4
testcase.c
testcase.c: In function 'f0a':
testcase.c:7:1: internal compiler error: in vec_safe_grow, at vec.h:618
 }
 ^
0xafa2c0 void vec_safe_grow<rtx_def const*, va_heap>(vec<rtx_def const*,
va_heap, vl_embed>*&, unsigned int)
        /mnt/svn/gcc-trunk/gcc/vec.h:618
0xafa2c0
generic_subrtx_iterator<const_rtx_accessor>::add_single_to_queue(generic_subrtx_iterator<const_rtx_accessor>::array_type&,
rtx_def const**, unsigned long, rtx_def const*)
        /mnt/svn/gcc-trunk/gcc/rtlanal.c:107
0xafa499
generic_subrtx_iterator<const_rtx_accessor>::add_subrtxes_to_queue(generic_subrtx_iterator<const_rtx_accessor>::array_type&,
rtx_def const**, unsigned long, rtx_def const*)
        /mnt/svn/gcc-trunk/gcc/rtlanal.c:174
0xe48b34 generic_subrtx_iterator<const_rtx_accessor>::next()
        /mnt/svn/gcc-trunk/gcc/rtl-iter.h:196
0xe48b34 ix86_loop_unroll_adjust
        /mnt/svn/gcc-trunk/gcc/config/i386/i386.c:51449
0x9cdcb3 decide_unroll_constant_iterations
        /mnt/svn/gcc-trunk/gcc/loop-unroll.c:403
0x9cdcb3 decide_unrolling
        /mnt/svn/gcc-trunk/gcc/loop-unroll.c:293
0x9cdcb3 unroll_loops(int)
        /mnt/svn/gcc-trunk/gcc/loop-unroll.c:311
0x9baa5f execute
        /mnt/svn/gcc-trunk/gcc/loop-init.c:611
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.


Tested revisions:
r222902 - ICE
5 r222437 - ICE
4_9 r222436 - OK

[Bug rtl-optimization/66076] [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[ubizjak at gmail dot com]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="UTF-8", Size: 12407 bytes --]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-05-10
                 CC|                            |rdsandiford at googlemail dot com,
                   |                            |ubizjak at gmail dot com
          Component|target                      |rtl-optimization
     Ever confirmed|0                           |1

--- Comment #1 from Uroš Bizjak <ubizjak at gmail dot com> ---
Confirmed.

#0  internal_error (gmsgid=gmsgid@entry=0x1687d37 "in %s, at %s:%d") at
/home/uros/gcc-svn/trunk/gcc/diagnostic.c:1217
#1  0x00000000011228d4 in fancy_abort (file=file@entry=0x1200878
"/home/uros/gcc-svn/trunk/gcc/vec.h", line=line@entry=618, 
    function=function@entry=0x12a8d80 <void vec_safe_grow<rtx_def const*,
va_heap>(vec<rtx_def const*, va_heap, vl_embed>*&, unsigned int)::__FUNCTION__>
"vec_safe_grow") at /home/uros/gcc-svn/trunk/gcc/diagnostic.c:1291
#2  0x0000000000a747e1 in vec_safe_grow<rtx_def const*, va_heap> (len=17,
v=@0x7fffffffd830: 0x1c84950)
    at /home/uros/gcc-svn/trunk/gcc/vec.h:618
#3  generic_subrtx_iterator<const_rtx_accessor>::add_single_to_queue
(array=..., base=0x7fffffffd7b0, i=i@entry=16, x=<optimized out>)
    at /home/uros/gcc-svn/trunk/gcc/rtlanal.c:107
#4  0x0000000000a7498a in
generic_subrtx_iterator<const_rtx_accessor>::add_subrtxes_to_queue (array=...,
base=<optimized out>, 
    base@entry=0x7fffffffd7b0, end=17, end@entry=1, x=0x7ffff1a1f010) at
/home/uros/gcc-svn/trunk/gcc/rtlanal.c:174
#5  0x0000000000daa055 in next (this=<synthetic pointer>) at
/home/uros/gcc-svn/trunk/gcc/rtl-iter.h:196
#6  ix86_loop_unroll_adjust (nunroll=1, loop=0x7ffff19cb510) at
/home/uros/gcc-svn/trunk/gcc/config/i386/i386.c:51449
#7  0x000000000097f367 in decide_unroll_constant_iterations (flags=<optimized
out>, loop=0x7ffff19cb510)
    at /home/uros/gcc-svn/trunk/gcc/loop-unroll.c:403
#8  decide_unrolling (flags=<optimized out>) at
/home/uros/gcc-svn/trunk/gcc/loop-unroll.c:293
#9  unroll_loops (flags=<optimized out>) at
/home/uros/gcc-svn/trunk/gcc/loop-unroll.c:311

While following patch should be used, it doesn't help. It looks that some
internal limit in RTX iterator is triggered for quite big RTX.

--cut here--
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c  (revision 222981)
+++ config/i386/i386.c  (working copy)
@@ -51446,7 +51446,7 @@ ix86_loop_unroll_adjust (unsigned nunroll, struct
   for (i = 0; i < loop->num_nodes; i++)
     FOR_BB_INSNS (bbs[i], insn)
       if (NONDEBUG_INSN_P (insn))
-       FOR_EACH_SUBRTX (iter, array, insn, NONCONST)
+       FOR_EACH_SUBRTX (iter, array, PATTERN (insn), NONCONST)
          if (const_rtx x = *iter)
            if (MEM_P (x))
              {
--cut here--

(gdb) f 6
#6  ix86_loop_unroll_adjust (nunroll=1, loop=0x7ffff19cb510) at
/home/uros/gcc-svn/trunk/gcc/config/i386/i386.c:51449
51449           FOR_EACH_SUBRTX (iter, array, PATTERN (insn), NONCONST)
(gdb) p debug_rtx (insn)
(insn 120 119 123 5 (set (reg:V32QI 230)
        (vec_select:V32QI (vec_concat:V64QI (subreg:V32QI (reg:V4DI 225) 0)
                (subreg:V32QI (reg:V4DI 225) 0))
            (parallel [
                    (const_int 8 [0x8])
                    (const_int 40 [0x28])
                    (const_int 9 [0x9])
                    (const_int 41 [0x29])
                    (const_int 10 [0xa])
                    (const_int 42 [0x2a])
                    (const_int 11 [0xb])
                    (const_int 43 [0x2b])
                    (const_int 12 [0xc])
                    (const_int 44 [0x2c])
                    (const_int 13 [0xd])
                    (const_int 45 [0x2d])
                    (const_int 14 [0xe])
                    (const_int 46 [0x2e])
                    (const_int 15 [0xf])
                    (const_int 47 [0x2f])
                    (const_int 24 [0x18])
                    (const_int 56 [0x38])
                    (const_int 25 [0x19])
                    (const_int 57 [0x39])
                    (const_int 26 [0x1a])
                    (const_int 58 [0x3a])
                    (const_int 27 [0x1b])
                    (const_int 59 [0x3b])
                    (const_int 28 [0x1c])
                    (const_int 60 [0x3c])
                    (const_int 29 [0x1d])
                    (const_int 61 [0x3d])
                    (const_int 30 [0x1e])
                    (const_int 62 [0x3e])
                    (const_int 31 [0x1f])
                    (const_int 63 [0x3f])
                ]))) pr66076.c:6 3549 {avx2_interleave_highv32qi}
     (expr_list:REG_DEAD (reg:V4DI 225)
        (nil)))
$15 = void

Adding RTX expert to CC.
>From gcc-bugs-return-485950-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sun May 10 16:26:04 2015
Return-Path: <gcc-bugs-return-485950-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 60401 invoked by alias); 10 May 2015 16:26:03 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 60370 invoked by uid 48); 10 May 2015 16:26:00 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug rtl-optimization/66076] [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march½ver4
Date: Sun, 10 May 2015 16:26:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: rtl-optimization
X-Bugzilla-Version: 6.0
X-Bugzilla-Keywords: ice-on-valid-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 5.2
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields: target_milestone
Message-ID: <bug-66076-4-8Bh1acZ6JB@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-66076-4@http.gcc.gnu.org/bugzilla/>
References: <bug-66076-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-05/txt/msg00790.txt.bz2
Content-length: 293

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |5.2
>From gcc-bugs-return-485951-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sun May 10 17:32:10 2015
Return-Path: <gcc-bugs-return-485951-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 62013 invoked by alias); 10 May 2015 17:32:10 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 61937 invoked by uid 48); 10 May 2015 17:32:06 -0000
From: "juergen.reuter at desy dot de" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug fortran/58586] ICE with derived type with allocatable component passed by value
Date: Sun, 10 May 2015 17:32:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: fortran
X-Bugzilla-Version: 4.8.0
X-Bugzilla-Keywords: ice-on-valid-code
X-Bugzilla-Severity: normal
X-Bugzilla-Who: juergen.reuter at desy dot de
X-Bugzilla-Status: WAITING
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: vehre at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-58586-4-I5WQTNFr7u@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-58586-4@http.gcc.gnu.org/bugzilla/>
References: <bug-58586-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-05/txt/msg00791.txt.bz2
Content-length: 379

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58586

--- Comment #5 from Jürgen Reuter <juergen.reuter at desy dot de> ---
Contrary to Dominique's comment in 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65894#c30
adding the patch in
https://gcc.gnu.org/ml/fortran/2015-04/msg00058.html
on top of r222970 doesn't break our code, and all of our tests run as intended.
>From gcc-bugs-return-485952-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Sun May 10 18:09:10 2015
Return-Path: <gcc-bugs-return-485952-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 115769 invoked by alias); 10 May 2015 18:09:10 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 115727 invoked by uid 55); 10 May 2015 18:09:06 -0000
From: "tkoenig at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug fortran/66041] [6 Regression] Matmul ICE
Date: Sun, 10 May 2015 18:09:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: fortran
X-Bugzilla-Version: 6.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: tkoenig at gcc dot gnu.org
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution:
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: tkoenig at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 6.0
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-66041-4-jUo9pIwpoR@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-66041-4@http.gcc.gnu.org/bugzilla/>
References: <bug-66041-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2015-05/txt/msg00792.txt.bz2
Content-length: 1144

https://gcc.gnu.org/bugzilla/show_bug.cgi?idf041

--- Comment #5 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
Author: tkoenig
Date: Sun May 10 18:08:33 2015
New Revision: 222982

URL: https://gcc.gnu.org/viewcvs?rev"2982&root=gcc&view=rev
Log:
2015-05-10  Thomas Koenig  <tkoenig@gcc.gnu.org>

        PR fortran/66041
        * frontend-passes.c (scalarized_expr): Set correct dimension and
        shape for the expression to be passed to lbound. Remove trailing
        references after array refrence.
        (inline_matmul_assign):  Remove gfc_copy_expr from calls
        to scalarized_expr().

2015-05-10  Thomas Koenig  <tkoenig@gcc.gnu.org>

        PR fortran/66041
        * gfortran.dg/inline_matmul_7.f90:  New test.
        * gfortran.dg/inline_matmul_8.f90:  New test.
        * gfortran.dg/inline_matmul_9.f90:  New test.


Added:
    trunk/gcc/testsuite/gfortran.dg/inline_matmul_7.f90
    trunk/gcc/testsuite/gfortran.dg/inline_matmul_8.f90
    trunk/gcc/testsuite/gfortran.dg/inline_matmul_9.f90
Modified:
    trunk/gcc/fortran/ChangeLog
    trunk/gcc/fortran/frontend-passes.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/66076] [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |rsandifo at gcc dot gnu.org
           Assignee|unassigned at gcc dot gnu.org      |rsandifo at gcc dot gnu.org

--- Comment #2 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
The bug occurs when we reuse an array from a previous
FOR_EACH_SUBRTX (which is supposed to be supported)
in cases where both iterations blow the stack array and
require the heap.  When moving from the stack to the heap
array the second time, we try to "grow" the heap array
to a size that's smaller than its current size.

Testing a fix.

[Bug rtl-optimization/66076] [5/6 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

--- Comment #3 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
Author: rsandifo
Date: Mon May 11 09:35:53 2015
New Revision: 222999

URL: https://gcc.gnu.org/viewcvs?rev=222999&root=gcc&view=rev
Log:
gcc/
        PR rtl-optimization/66076
        * rtlanal.c (generic_subrtx_iterator <T>::add_single_to_queue):
        Don't grow the heap array if it is already big enough from a
        previous iteration.

gcc/testsuite/
        PR rtl-optimization/66076
        * gcc.dg/torture/pr66076.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr66076.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/rtlanal.c
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/66076] [5 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

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

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

[Bug rtl-optimization/66076] [5 Regression] ICE: in vec_safe_grow, at vec.h:618 with -funroll-loops -mno-prefer-avx128 -march=bdver4

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66076

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|5.2                         |5.3

--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 5.2 is being released, adjusting target milestone to 5.3.