[Bug bootstrap/102527] New: [12 regression] out of memory compiling insn-emit.c

[Bug bootstrap/102527] New: [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

            Bug ID: 102527
           Summary: [12 regression] out of memory compiling insn-emit.c
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ro at gcc dot gnu.org
                CC: aldyh at gcc dot gnu.org
  Target Milestone: ---
              Host: i386-pc-solaris2.11
            Target: i386-pc-solaris2.11
             Build: i386-pc-solaris2.11

Between 20210927 (rev dd11aab6463880c35d942c4a4fd346fdaeeb8e72) and
20210928 (rev a11052d98db2f2a61841f0c5ee84de4ca1b3e296), 32-bit Solaris/x86
bootstrap broke compiling insn-emit.c in stage 2:

cc1plus: out of memory allocating 65536 bytes after a total of 2747670528 bytes
make[3]: *** [Makefile:1136: insn-emit.o] Error 1

Comparing to the cc1plus memory consumption in the working configuration,
memory usage has almost quadrupled from ca. 750 MB to ca. 2.6 GB.

A reghunt identified this patch as the culprit:

commit 0288527f47cec6698b31ccb3210816415506009e
Author: Aldy Hernandez <aldyh@redhat.com>
Date:   Tue Sep 21 10:27:53 2021 +0200

    Replace VRP threader with a hybrid forward threader.

sparc-sun-solaris2.11 isn't affected, neither is i686-pc-linux-gnu.

Compiling with -fmem-report doesn't produce anything in the failing case.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

Rainer Orth <ro at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |12.0

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

David Edelsohn <dje at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dje at gcc dot gnu.org
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-09-29
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from David Edelsohn <dje at gcc dot gnu.org> ---
Confirmed.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #2 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to David Edelsohn from comment #1)
> Confirmed.

I don't have access to an i386-solaris box.

David, how were you able to reproduce?

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #2 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> (In reply to David Edelsohn from comment #1)
>> Confirmed.
>
> I don't have access to an i386-solaris box.

... and there's none in the cfarm unfortunately, and sparc-solaris is
unaffected.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #4 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #3)
> > --- Comment #2 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> > (In reply to David Edelsohn from comment #1)
> >> Confirmed.
> >
> > I don't have access to an i386-solaris box.
> 
> ... and there's none in the cfarm unfortunately, and sparc-solaris is
> unaffected.

Ughh..

Can it be reproduced on a cross?  If so, could you post the .ii file with your
include file peculiarities?

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #5 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Created attachment 51518
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51518&action=edit
patch to help diagnose issue with -ftime-report

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #6 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
The attached patch adds a separate TV_* timer to see the actual break down for
VRP and the VRP threader.

Could you incorporate this patch and run the problematic file with ./cc1
-ftime-report -O2?  I'd like to see the table at the end.  If you're so
inclined, could you also email me the entire -ftime-report so I can see the GC
output as it compiles?

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #7 from Rainer Orth <ro at gcc dot gnu.org> ---
(In reply to Aldy Hernandez from comment #4)

> Can it be reproduced on a cross?  If so, could you post the .ii file with
> your include file peculiarities?

I haven't tried, but expect it should work, provided you run on a 32-bit host
and perhaps limit VM consumption.  I expect that depending on the host OSes
VM layout cc1plus may be able to use more of the 4 GB address space on one OS
than the other.

That said, there's a weird problem:

If I run a regular build, I get the `out of memory' error reported.  Just
adding
-save-temps to the command line produces a couple of weird warnings/errors:

/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_495(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_496(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_497(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_498(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_499(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_500(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_501(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md: In function ‘rtx_insn*
gen_split_502(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20130:29: error: this statement
may fall through [-Werror=implicit-fallthrough=]
20130 |       code = swap_condition (code);
      |              ~~~~~~~~~~~~~~~^~~~~~
/var/gcc/reghunt/master/gcc/config/i386/i386.md:20133:5: note: here
20133 |     case GE: case GEU:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/sse.md: In function ‘rtx_insn*
gen_split_2186(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/sse.md:18256:19: error: this statement
may fall through [-Werror=implicit-fallthrough=]
18256 |       operands[1] = tmp;
      |       ~~~~~~~~~~~~^~~~~
/var/gcc/reghunt/master/gcc/config/i386/sse.md:18258:5: note: here
18258 |     case 32:
      |     ^~~~
/var/gcc/reghunt/master/gcc/config/i386/sse.md: In function ‘rtx_insn*
gen_split_2187(rtx_insn*, rtx_def**)’:
/var/gcc/reghunt/master/gcc/config/i386/sse.md:18256:19: error: this statement
may fall through [-Werror=implicit-fallthrough=]
18256 |       operands[1] = tmp;
      |       ~~~~~~~~~~~~^~~~~
/var/gcc/reghunt/master/gcc/config/i386/sse.md:18258:5: note: here
18258 |     case 32:
      |     ^~~~

and (if adding -Wno-error to avoid that) while cc1plus still consumes in excess
of 3 GB VM space, the compilation finally succeeds.

I'm attaching the preprocessed output nonetheless; maybe it's still helpful.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #8 from Rainer Orth <ro at gcc dot gnu.org> ---
Created attachment 51521
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51521&action=edit
i386-pc-solaris2.11 insn-emit.ii

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #9 from Rainer Orth <ro at gcc dot gnu.org> ---
Created attachment 51522
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51522&action=edit
insn-emit.ii -ftime-report output

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #10 from Rainer Orth <ro at gcc dot gnu.org> ---
(In reply to Aldy Hernandez from comment #6)
> The attached patch adds a separate TV_* timer to see the actual break down
> for VRP and the VRP threader.
> 
> Could you incorporate this patch and run the problematic file with ./cc1
> -ftime-report -O2?  I'd like to see the table at the end.  If you're so
> inclined, could you also email me the entire -ftime-report so I can see the
> GC output as it compiles?

Sure: I've attached the output of

cc1plus -fpreprocessed insn-emit.ii -quiet -mtune=generic -march=pentium4 -g
-O2 -o insn-emit.s -ftime-report

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at redhat dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #11 from Aldy Hernandez <aldyh at redhat dot com> ---
This looks mighty suspicious ;-)

diff --git a/gcc/tree-vrp.c b/gcc/tree-vrp.c
index 69a3ab0ea9d..c24c67f8874 100644
--- a/gcc/tree-vrp.c
+++ b/gcc/tree-vrp.c
@@ -4408,6 +4408,7 @@ hybrid_threader::~hybrid_threader ()
   delete m_threader;
   delete m_state;
   delete m_ranger;
+  delete m_query;

   scev_finalize ();
   loop_optimizer_finalize ();

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #12 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to Aldy Hernandez from comment #11)
> This looks mighty suspicious ;-)
> 
> diff --git a/gcc/tree-vrp.c b/gcc/tree-vrp.c
> index 69a3ab0ea9d..c24c67f8874 100644
> --- a/gcc/tree-vrp.c
> +++ b/gcc/tree-vrp.c
> @@ -4408,6 +4408,7 @@ hybrid_threader::~hybrid_threader ()
>    delete m_threader;
>    delete m_state;
>    delete m_ranger;
> +  delete m_query;
> 
>    scev_finalize ();
>    loop_optimizer_finalize ();

FWIW, all the patches so far proposed have been pushed to trunk, if you want to
try again.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

--- Comment #13 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #12 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> (In reply to Aldy Hernandez from comment #11)
>> This looks mighty suspicious ;-)
>> 
>> diff --git a/gcc/tree-vrp.c b/gcc/tree-vrp.c
>> index 69a3ab0ea9d..c24c67f8874 100644
>> --- a/gcc/tree-vrp.c
>> +++ b/gcc/tree-vrp.c
>> @@ -4408,6 +4408,7 @@ hybrid_threader::~hybrid_threader ()
>>    delete m_threader;
>>    delete m_state;
>>    delete m_ranger;
>> +  delete m_query;
>> 
>>    scev_finalize ();
>>    loop_optimizer_finalize ();
>
> FWIW, all the patches so far proposed have been pushed to trunk, if you want to
> try again.

This has done the trick: current master bootstraps again on
i386-pc-solaris2.11.

Thanks.

[Bug bootstrap/102527] [12 regression] out of memory compiling insn-emit.c

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102527

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

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

--- Comment #14 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
fixed