[Bug c/39326] New: Segmentation fault with -O1, out of memory with -O2

[Bug c/39326] New: Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1325 bytes --]

I was trying to compile autogenerated by 'spiral' 'gap_TlnLv4.c' file (to be
attached).

Trying to compile it with

/mnt/sdb8/sergei/AFSWD_debug/install/gcc-4.3.3/binsh/gcc -O1
-fomit-frame-pointer -malign-double -fstrict-aliasing -march=native -c
/tmp/spiral-sergei/gap_TlnLv4.c -o gap_TlnLv4.o1

command line I got this:

"
/tmp/spiral-sergei/gap_TlnLv4.c: In function ‘RDFT_49152_1’:
/tmp/spiral-sergei/gap_TlnLv4.c:172282: internal compiler error: Segmentation
fault
Please submit a full bug report,
".

Before that, trying to compile it with

/mnt/sdb8/sergei/AFSWD_debug/install/gcc-4.3.3/binsh/gcc -O2
-fomit-frame-pointer -malign-double -fstrict-aliasing -march=native -c
/tmp/spiral-sergei/gap_TlnLv4.c

command I got this:

cc1: out of memory allocating 4184025948 bytes after a total of 205533184 bytes
.


-- 
           Summary: Segmentation fault with -O1, out of memory with -O2
           Product: gcc
           Version: 4.3.3
            Status: UNCONFIRMED
          Severity: major
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: sergstesh at yahoo dot com
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #1 from sergstesh at yahoo dot com  2009-02-28 15:29 -------
Created an attachment (id=17377)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17377&action=view)
source file causing the failure

The source is not preprocessed, it has only standard

#include <math.h>

in it.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #2 from sergstesh at yahoo dot com  2009-02-28 15:32 -------
My OS:

Linux amdam2 2.6.22.19-0.2-default #1 SMP 2008-12-18 10:17:03 +0100 i686 athlon
i386 GNU/Linux

- SUSE 10.3 in simple English; 'gcc' is self-built 'gcc-4.3.3'.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #3 from sergstesh at yahoo dot com  2009-02-28 15:34 -------
There is no failure with -O0.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #4 from sergstesh at yahoo dot com  2009-02-28 17:23 -------
FWIW, 'gcc-3.4.6', when run as

/mnt/sdb8/sergei/AFSWD_debug/install/gcc-3.4.6/binsh/gcc -O1
-fomit-frame-pointer -malign-double -fstrict-aliasing -c
/tmp/spiral-sergei/gap_TlnLv4.c -o gap_TlnLv4.o

, fails with

cc1: out of memory allocating 182853324 bytes after a total of 14716928 bytes

message, not with segmentation fault like 'gcc-4.3.3'.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #5 from rguenth at gcc dot gnu dot org  2009-02-28 17:57 -------
Try -fno-move-loop-invariants.  This is probably a killer on alias-improvements
branch as well.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu dot
                   |                            |org
           Keywords|                            |compile-time-hog, memory-hog


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #6 from rguenth at gcc dot gnu dot org  2009-02-28 18:00 -------
As this seems to be autogenerated from

! The SPL Program: (compose (sparse (coords (...12288 x 2 ...))(values (...1 x
12288 ...)))(compose (conjugate (..)(..))(compose (..)(..))))
! node size: 12288 X 12288

can you produce a testcase with an order of magnitude less loads/stores?
(thus likely just "node size: 1228 x 1228")?

Thanks!


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #7 from rguenth at gcc dot gnu dot org  2009-02-28 18:10 -------
One problem seems to be that we blow up the stack during GC recursing following
the VUSE -> def_stmt links.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #8 from sergstesh at yahoo dot com  2009-03-01 03:03 -------
Created an attachment (id=17378)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17378&action=view)
a smaller file with hopefully the same pattern


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #9 from sergstesh at yahoo dot com  2009-03-01 03:06 -------
(In reply to comment #6)
> As this seems to be autogenerated from
> 
> ! The SPL Program: (compose (sparse (coords (...12288 x 2 ...))(values (...1 x
> 12288 ...)))(compose (conjugate (..)(..))(compose (..)(..))))
> ! node size: 12288 X 12288
> 
> can you produce a testcase with an order of magnitude less loads/stores?
> (thus likely just "node size: 1228 x 1228")?
> 
> Thanks!
> 

I've uploaded a smaller file: gap_bzAJWH.c.gz . These are intermediated files
which are deleted upon successful compilation.

So, I've decreased the number the points and managed to "catch" this one. It
is: "node size: 1536 X 1536".


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #10 from sergstesh at yahoo dot com  2009-03-01 03:09 -------
I am not sure whether it's clear - the smaller 'gap_bzAJWH.c.gz' file can be
found as

http://gcc.gnu.org/bugzilla/attachment.cgi?id=17378&action=view

attachment.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 783 bytes --]



------- Comment #11 from sergstesh at yahoo dot com  2009-03-01 03:54 -------
(In reply to comment #5)
> Try -fno-move-loop-invariants.  This is probably a killer on alias-improvements
> branch as well.
> 

Still segfault:

"
gap_TlnLv4.c: In function ‘RDFT_49152_1’:
gap_TlnLv4.c:172282: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

real    33m52.152s
user    16m47.843s
sys     0m9.333s

[1]   Exit 1                  time
/mnt/sdb8/sergei/AFSWD_debug/install/gcc-4.3.3/binsh/gcc -O1
-fomit-frame-pointer -malign-double -fstrict-aliasing -fno-move-loop-invariants
-c gap_TlnLv4.c
".


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #12 from rguenth at gcc dot gnu dot org  2009-03-01 11:39 -------
The small testcase at -O2 is bound by PRE time

 PRE                   :  27.08 (55%) usr   0.02 ( 4%) sys  27.09 (55%) wall   
1551 kB ( 2%) ggc

nothing interesting at -O1 and it still nicely fits into my memory.  Which
makes me as for a little bigger testcase ;)  (as I suspect there's quadratic
behavior in memory usage somewhere a size around 2000 * 2000 would be nice)

Thanks again.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-03-01 11:39:34
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug c/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #13 from rguenth at gcc dot gnu dot org  2009-03-02 17:00 -------
And for trunk we have

 combiner              :  24.76 (15%) usr   0.80 (23%) sys  25.76 (16%) wall 
448053 kB (82%) ggc
 integrated RA         :  68.41 (43%) usr   1.06 (31%) sys  70.23 (42%) wall   
3197 kB ( 1%) ggc

as the worst offenders at -O1.  I notice that with --enable-checking GC
params the top memory usage is way lower than without, so something odd
happens here as well (maybe a ggc_free wrapped inside ENABLE_CHECKING?).

Most memory is used by combine and IRA.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|major                       |normal


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #14 from sergstesh at yahoo dot com  2009-03-03 13:36 -------
'spiral' has produced another testcase which segfaults with -O2 - the original
testcase segfaults with -O1.

The testcase, though has half the points if terms of FFT, is big as a file:

-rw-r--r-- 1 sergei users  1656419 2009-03-03 06:36 gap_CPEodL.c.bz2

, so I think I can't upload it directly, or can I ?

I can send it as Email attachment to, say, rguenth.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[rguenther at suse dot de]

------- Comment #15 from rguenther at suse dot de  2009-03-03 13:48 -------
Subject: Re:  Segmentation fault with -O1, out of memory
 with -O2

On Tue, 3 Mar 2009, sergstesh at yahoo dot com wrote:

> ------- Comment #14 from sergstesh at yahoo dot com  2009-03-03 13:36 -------
> 'spiral' has produced another testcase which segfaults with -O2 - the original
> testcase segfaults with -O1.
> 
> The testcase, though has half the points if terms of FFT, is big as a file:
> 
> -rw-r--r-- 1 sergei users  1656419 2009-03-03 06:36 gap_CPEodL.c.bz2
> 
> , so I think I can't upload it directly, or can I ?
> 
> I can send it as Email attachment to, say, rguenth.

That would be fine.

Richard.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[sergstesh at yahoo dot com]

------- Comment #16 from sergstesh at yahoo dot com  2009-03-03 14:15 -------
(In reply to comment #15)
> Subject: Re:  Segmentation fault with -O1, out of memory
>  with -O2
> 
> On Tue, 3 Mar 2009, sergstesh at yahoo dot com wrote:
> 
> > ------- Comment #14 from sergstesh at yahoo dot com  2009-03-03 13:36 -------
> > 'spiral' has produced another testcase which segfaults with -O2 - the original
> > testcase segfaults with -O1.
> > 
> > The testcase, though has half the points if terms of FFT, is big as a file:
> > 
> > -rw-r--r-- 1 sergei users  1656419 2009-03-03 06:36 gap_CPEodL.c.bz2
> > 
> > , so I think I can't upload it directly, or can I ?
> > 
> > I can send it as Email attachment to, say, rguenth.
> 
> That would be fine.
> 
> Richard.
> 

I have just sent Richard an Email with 'gap_CPEodL.c.bz2' file attached.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #17 from rguenth at gcc dot gnu dot org  2009-03-17 11:05 -------
Created an attachment (id=17476)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17476&action=view)
the other testcase


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #18 from rguenth at gcc dot gnu dot org  2009-03-17 12:42 -------
Vlad, for the second testcase I see

-O0:

 expand                :   0.78 (19%) usr   0.04 ( 5%) sys   0.83 (16%) wall  
44335 kB (49%) ggc
 integrated RA         :   1.05 (25%) usr   0.03 ( 4%) sys   1.13 (22%) wall   
1036 kB ( 1%) ggc
 reload                :   0.55 (13%) usr   0.02 ( 3%) sys   0.59 (12%) wall   
3585 kB ( 4%) ggc

-O1:

 integrated RA         :  73.95 (77%) usr   1.18 (50%) sys  76.20 (76%) wall   
2898 kB ( 3%) ggc

(on alias improvements branch I see
 combiner              :  24.89 (17%) usr   0.51 (28%) sys  25.41 (17%) wall 
448053 kB (82%) ggc
 integrated RA         :  69.81 (47%) usr   0.36 (20%) sys  70.22 (46%) wall   
3212 kB ( 1%) ggc
as tree loop invariant motion does lots of stuff there)

-O2:

 PRE                   :  25.49 (26%) usr   0.03 ( 3%) sys  25.52 (26%) wall   
3513 kB ( 3%) ggc
 integrated RA         :  36.98 (38%) usr   0.24 (21%) sys  37.31 (38%) wall   
2158 kB ( 2%) ggc


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vmakarov at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #19 from rguenth at gcc dot gnu dot org  2009-03-17 12:58 -------
For trunk -O1 I see

CPU: AMD64 processors, speed 1000 MHz (estimated)
Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask
of 0x00 (No unit mask) count 100000
samples  %        symbol name
368029   19.8218  allocno_spill_priority_compare
244104   13.1473  color_pass
205725   11.0802  build_allocno_conflicts
141509    7.6216  push_allocno_to_stack
106750    5.7495  assign_hard_reg
95558     5.1467  bitmap_bit_p
88785     4.7819  add_to_allocno_conflicts
51962     2.7986  splay_tree_splay
42764     2.3032  ira_build_conflicts
40718     2.1930  ira_flattening
27977     1.5068  canon_true_dependence
25608     1.3792  ira_allocno_live_ranges_intersect_p
20403     1.0989  nonoverlapping_memrefs_p
20005     1.0775  ira_add_allocno_conflict


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326

[Bug middle-end/39326] Segmentation fault with -O1, out of memory with -O2

[rguenth at gcc dot gnu dot org]

------- Comment #20 from rguenth at gcc dot gnu dot org  2009-03-17 13:09 -------
Btw, it looks like internal IRA data-structures can be shrinked on 64bit
platforms by avoiding padding between pointer and integer members a lot.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39326