Re: preprocessor/10323: The preprocessor consumes all the memory

Re: preprocessor/10323: The preprocessor consumes all the memory

[Neil Booth]

The following reply was made to PR preprocessor/10323; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: Zack Weinberg <zack@codesourcery.com>
Cc: lu_zero@gentoo.org, gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/10323: The preprocessor consumes all the memory
Date: Mon, 7 Apr 2003 07:07:23 +0100

 Zack Weinberg wrote:-
 
 > lu_zero@gentoo.org writes:
 > 
 > > Preprocessing altivec code it just consumes all the memory and then
 > > dies reporting:
 > > cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
 > 
 > I can confirm this with mainline, it looks like infinite recursion in
 > the macro expander.  Neil, can you take a look?
 
 I'm pretty certain that the macro expander doesn't recurse or loop.
 
 I've looked at it briefly.  It looks like everything is OK - the expansion
 is just 100s of megabytes or even gigabytes big.
 
 The line that dies is (you can cut the main file down to just this line)
 
     tmp = vec_sub (vec_avg (avg0, avg1),
     		   vec_and (vec_and (ones, vec_or (xor0, xor1)),
 		   			    vec_xor (avg0, avg1)));
 
 Just the 2nd line here by itself expands to almost 6 megs.
 I think it's fair to expect the whole lot to be "very big".
 
 Can you get any other compiler to compile it?  If you can, I'll look at
 it more closely, but I think it's a case of "don't do that, use 
 temporaries instead".
 
 Neil.

Re: preprocessor/10323: The preprocessor consumes all the memory

[neil]

Synopsis: The preprocessor consumes all the memory

State-Changed-From-To: open->closed
State-Changed-By: neil
State-Changed-When: Mon Apr  7 07:15:45 2003
State-Changed-Why:
    Not a bug.  The macro expansion really is that big.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=10323

Re: preprocessor/10323: The preprocessor consumes all the memory

[Zack Weinberg]

The following reply was made to PR preprocessor/10323; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: Neil Booth <neil@daikokuya.co.uk>
Cc: lu_zero@gentoo.org,  gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/10323: The preprocessor consumes all the memory
Date: Sun, 06 Apr 2003 23:36:31 -0700

 Neil Booth <neil@daikokuya.co.uk> writes:
 
 > Zack Weinberg wrote:-
 >
 >> lu_zero@gentoo.org writes:
 >> 
 >> > Preprocessing altivec code it just consumes all the memory and then
 >> > dies reporting:
 >> > cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
 >> 
 >> I can confirm this with mainline, it looks like infinite recursion in
 >> the macro expander.  Neil, can you take a look?
 >
 > I'm pretty certain that the macro expander doesn't recurse or loop.
 >
 > I've looked at it briefly.  It looks like everything is OK - the expansion
 > is just 100s of megabytes or even gigabytes big.
 
 Oh, ok, I probably just got sick of waiting and ^C-ed it before it
 finished / ran out of memory.
 
 zw

Re: preprocessor/10323: The preprocessor consumes all the memory

[Zack Weinberg]

The following reply was made to PR preprocessor/10323; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: Neil Booth <neil@daikokuya.co.uk>
Cc: lu_zero@gentoo.org,  gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/10323: The preprocessor consumes all the memory
Date: Sun, 06 Apr 2003 15:59:28 -0700

 Neil Booth <neil@daikokuya.co.uk> writes:
 
 > Can someone give me altivec.h?  I hope it doesn't include anything else.
 
 gcc/config/rs6000/altivec.h
 
 zw

Re: preprocessor/10323: The preprocessor consumes all the memory

[Neil Booth]

The following reply was made to PR preprocessor/10323; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: Zack Weinberg <zack@codesourcery.com>
Cc: lu_zero@gentoo.org, gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/10323: The preprocessor consumes all the memory
Date: Sun, 6 Apr 2003 23:55:13 +0100

 Zack Weinberg wrote:-
 
 > lu_zero@gentoo.org writes:
 > 
 > > Preprocessing altivec code it just consumes all the memory and then
 > > dies reporting:
 > > cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
 > 
 > I can confirm this with mainline, it looks like infinite recursion in
 > the macro expander.  Neil, can you take a look?
 
 Can someone give me altivec.h?  I hope it doesn't include anything else.
 
 Neil.

Re: preprocessor/10323: The preprocessor consumes all the memory

[Zack Weinberg]

The following reply was made to PR preprocessor/10323; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: lu_zero@gentoo.org, Neil Booth <neil@daikokuya.co.uk>
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/10323: The preprocessor consumes all the memory
Date: Sun, 06 Apr 2003 15:36:11 -0700

 lu_zero@gentoo.org writes:
 
 > Preprocessing altivec code it just consumes all the memory and then
 > dies reporting:
 > cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
 
 I can confirm this with mainline, it looks like infinite recursion in
 the macro expander.  Neil, can you take a look?
 
 zw

preprocessor/10323: The preprocessor consumes all the memory

[lu_zero]

>Number:         10323
>Category:       preprocessor
>Synopsis:       The preprocessor consumes all the memory
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Apr 06 21:26:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     lu_zero@gentoo.org
>Release:        gcc 3.2.2
>Organization:
>Environment:
powerpc-unknown-linux-gnu 
PowerPC 7455
512mb ram
linux-2.4.20-benh9
glibc-2.3.2
>Description:
Preprocessing altivec code it just consumes all the memory and then dies reporting:                                                                                
cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
the system has 512mb ram and 800mb swap
The file is a cut down motion_comp_altivec.c from libmpeg2, that I just used to restrict the faulting part (the rest is compiling ok)
>How-To-Repeat:
that is enough
gcc -c -Os -maltivec -mabi=altivec -D_REENTRANT -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -I. -I../libvo -I..   -DMPG12PLAY -o motion_comp_altivec.2.o motion_comp_altivec.2.c -E
                                                                                
that is the result
cpp0: Cannot allocate 262144112 bytes after allocating 7222368 bytes
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: text/x-csrc; name="motion_comp_altivec.2.c"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="motion_comp_altivec.2.c"
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