Discussions about the Boehm Garbage Collector (Boehm GC).

Ivan Maidanski | 4 Nov 2011 13:58

Picon

Re[2]: [Gc] weak maps and libgc

Hi Andy,

Look good to me. I've checked it into my working tree (and commit in several days I hope).
I've done one change to the proposed function - in case dest==src, it now returns success (instead of GC_DUPLICATE).

Also, I've added some tests regarding the function.

Regards.

26 10 2011, 20:18 Andy Wingo <wingo@...>:
> On Wed 26 Oct 2011 10:18, Ivan Maidanski <ivmai@...> writes:
> 
> >>  (2) a function to move disappearing links, perhaps callable only
> >>      within the alloc lock.
> >> 
> >>      I did some profiling today with the open-addressed table.  It has
> >>      to shuffle items around sometimes.  Unregistering then registering
> >>      disappearing links shows up pretty high on the profile.
> >
> > Is my understanding correct - you want to optimize the following:
> >
> > // assume GC_general_register_disappearing_link(&link, obj1);
> > ...
> > GC_unregister_disappearing_link(&link);
> > GC_general_register_disappearing_link(&link, obj2);
> >
> > Right? If yes, I don't mind again.
> 
> Patch attached.  WDYT?
>

(Continue reading)

Boehm, Hans | 8 Nov 2011 02:23

Picon

RE: Re[4]: weak maps and libgc

Sorry about the belated reply.  Again.

> -----Original Message-----
> From: Ivan Maidanski [mailto:ivmai@...]
> Sent: Wednesday, October 26, 2011 2:36 AM
> To: Boehm, Hans; Andy Wingo
> Cc: gc
> Subject: Re[4]: [Gc] weak maps and libgc
> 
> Hi Hans,
> 
> What's your opinion of these 3 ideas?
> 
> PS. Anyway, it won't go into 7.2 branch.
> 
> Regards.
> 
> 26 10 2011, 13:05 Andy Wingo <wingo@...>:
> > Thanks for the thoughts, Ivan!
> >
> > On Wed 26 Oct 2011 10:18, Ivan Maidanski <ivmai@...> writes:
> >
> > > 26 10 2011, 00:06 Andy Wingo <wingo@...>:
> > >>  (1) a variant of GC_register_disappearing_link that also
> > >>      incremented an `unsigned long' somewhere in memory
> > >
> > > You mean statistic counter, right?
> > > Good idea.
> >
> > Yes, but ideally one that can be associated with a particular memory

(Continue reading)

Ivan Maidanski | 8 Nov 2011 08:23

Picon

Re[6]: [Gc] weak maps and libgc

Hi Hans,

Regarding GC_move_disappearing_link: I pushed the code to master yesterday (finalize.c).
To care about app code size, I think it's better to move all disappearing-link-related functions to a
standalone .c file.
BTW: GC_move_disappearing_link(&link,&link) can be use to test whether link is registered.

Regards.

08 11 2011, 05:29 "Boehm, Hans" <hans.boehm@...>:
> Sorry about the belated reply.  Again.
> 
> > -----Original Message-----
> > From: Ivan Maidanski [mailto:ivmai@...]
> > Sent: Wednesday, October 26, 2011 2:36 AM
> > To: Boehm, Hans; Andy Wingo
> > Cc: gc
> > Subject: Re[4]: [Gc] weak maps and libgc
> >
> > Hi Hans,
> >
> > What's your opinion of these 3 ideas?
> >
> > PS. Anyway, it won't go into 7.2 branch.
> >
> > Regards.
> >
> > 26 10 2011, 13:05 Andy Wingo <wingo@...>:
> > > Thanks for the thoughts, Ivan!
> > >

(Continue reading)

Ivan Maidanski | 9 Nov 2011 11:56

Picon

Re[7]: [Gc] weak maps and libgc

Hi Hans and Andy,

08 11 2011, 11:23 Ivan Maidanski <ivmai@...>:
> Hi Hans,
> 
> Regarding GC_move_disappearing_link: I pushed the code to master yesterday (finalize.c).
> To care about app code size, I think it's better to move all disappearing-link-related functions to a
standalone .c file.

Sorry, I was wrong about this (as disappearing-link-related code is invoked from other parts of GC and
moving only its API functions to a standalone file would require to make some global vars non-static).

So, I've added a new macro to prohibit compilation of that function on demand -
GC_MOVE_DISAPPEARING_LINK_NOT_NEEDED (similar to JAVA_FINALIZATION_NOT_NEEDED).

Regards.

> BTW: GC_move_disappearing_link(&link,&link) can be use to test whether link is registered.
> 
> Regards.
> 
> 08 11 2011, 05:29 "Boehm, Hans" <hans.boehm@...>:
> > Sorry about the belated reply.  Again.
> >
> > > -----Original Message-----
> > > From: Ivan Maidanski [mailto:ivmai@...]
> > > Sent: Wednesday, October 26, 2011 2:36 AM
> > > To: Boehm, Hans; Andy Wingo
> > > Cc: gc
> > > Subject: Re[4]: [Gc] weak maps and libgc

(Continue reading)

Johann 'Myrkraverk' Oskarsson | 15 Nov 2011 00:33

Why does ./configure check for g77?

Hi all,

Why does the configure script check for g77?  Is there fortran support
in the collector?

checking for g77... g77
checking whether we are using the GNU Fortran 77 compiler... yes
checking whether g77 accepts -g... yes

checking for g77 option to produce PIC... -fPIC
checking if g77 PIC flag -fPIC works... yes
checking if g77 static flag -static works... no
checking if g77 supports -c -o file.o... yes
checking whether the g77 linker (/usr/ccs/bin/ld) supports shared
libraries... yes

--

-- 
   Johann Oskarsson                http://www.2ndquadrant.com/    |[]
   PostgreSQL Development, 24x7 Support, Training and Services  --+--
                                                                  |
   Blog: http://my.opera.com/myrkraverk/blog/

Andy Wingo | 15 Nov 2011 11:38

Picon

Re: Why does ./configure check for g77?

On Tue 15 Nov 2011 00:33, "Johann 'Myrkraverk' Oskarsson"
<johann@...> writes:

> Why does the configure script check for g77?  Is there fortran support
> in the collector?

Probably something related to libtool.  See "Tags" in the libtool
manual.

Andy
--

-- 
http://wingolog.org/

Ivan Maidanski | 16 Nov 2011 17:33

Picon

Re[3]: [Gc] [PATCH 2/2]: Hexagon atomic ops

Hi Linas,

Based on your implementation of CAS for Hexagon, I've created another CAS primitive equivalent to GCC
__sync_val_compare_and_swap one (see AO_fetch_compare_and_swap() in https://github.com/ivmai/libatomic_ops/blob/ao_val_cas/src/atomic_ops/sysdeps/gcc/hexagon.h).
But I have no opportunity to test the code.

Could you please test that code ("ao_val_cas" branch) for Hexagon?

git clone git://github.com/ivmai/libatomic_ops.git -b ao_val_cas
cd libatomic_ops; autoreconf -vif
./configure; make check

Thanks.

17 10 2011, 22:18 Ivan Maidanski <ivmai@...>:
> Hi Linas,
> 
> I've merged this into 'release' (v7.2) branch as well.
> 
> Regards.
> 
> 12 10 2011, 21:39 Linas Vepstas <linasvepstas@...>:
> > On 11 October 2011 00:26, Ivan Maidanski <ivmai@...> wrote:
> >
> > > I've committed it as patch 9136ee6 to libatomic_ops repo master branch with minor changes (in
particular, I've changed the words in the license to match that on other files - the sense should remain the
same as in the original file you sent but please check it).
> >
> > Looks reasonable. I had cut-n-pasted the license notice from some
> > other file in that directory, with the goal of making it compatible.

(Continue reading)

Ivan Maidanski | 16 Nov 2011 18:07

Picon

sync_val_compare_and_swap equivalent in libatomic_ops

Hi all,

I've added CAS primitive which returns old value instead of comparison result (the similar primitives in
GCC are __sync_val_compare_and_swap and __sync_bool_compare_and_swap, respectively). I've named
the primitive as AO_fetch_compare_and_swap.

I've added implementations for emulations, x86, x64, arm v6, Hexagon (not tested), ia64 (trivial but not
tested), ppc (not tested) targets.
Meanwhile, alpha, avr, m68k, mips, s390, sparc lack such CAS primitive for now (unless via emulation) - I'm
unable to do this without help from the community.

For some targets (e.g., ia64), old bool CAS primitives containing "return oldval_fetched ==
oldval_expected" statement are completely removed from "sysdeps" part as they are implemented in
generalize.h (and generalize-small.h).

The functionality is available in ao_val_cas branch for testing period.
For implementation details, see the differences from master branch - https://github.com/ivmai/libatomic_ops/compare/master...ao_val_cas

Tests for new CAS has been added, so please everyone is welcomed to participate in testing:

git clone git://github.com/ivmai/libatomic_ops.git -b ao_val_cas
cd libatomic_ops
autoreconf -vif
./configure
make check

Regards.

Ivan Maidanski | 16 Nov 2011 18:20

Picon

Re[2]: [Gc] Re: avr32 port patch (from Debian)

Hi Bradley,

Could you also please supply a CAS primitive returning old value (instead of the comparison result)?

I've added such primitive e.g., for x86, arm - as a reference, you could see AO_fetch_compare_and_swap() 
implementation in:
https://github.com/ivmai/libatomic_ops/blob/ao_val_cas/src/atomic_ops/sysdeps/generic_pthread.h
https://github.com/ivmai/libatomic_ops/blob/ao_val_cas/src/atomic_ops/sysdeps/gcc/arm.h
https://github.com/ivmai/libatomic_ops/blob/ao_val_cas/src/atomic_ops/sysdeps/gcc/x86.h

The code is currently available only in ao_val_cas feature branch:
git clone git://github.com/ivmai/libatomic_ops.git -b ao_val_cas

Thanks.

30 05 2010, 19:07 Ivan Maidanski <ivmai@...>:
> 
> Sun, 30 May 2010 16:26:27 +0200 Stefano Rivera <stefano@...>:
> 
> > Hi gc (2010.05.20_22:13:14_+0200)
> > > Debian has been carrying a patch porting libgc to avr32 since August
> > > 2009 [0], which as far as I know has never been forwarded to you.
> >
> > Whoops, looks like that patch was incomplete [0].
> >
> > Here's an updated version [1] from the Debian avr32 porter.
> 
> I've committed the patch with the following changes:
> 1. AO_compare_and_swap -> AO_compare_and_swap_full;
> 2. "unsigned long int" -> "long";

(Continue reading)

Linas Vepstas | 16 Nov 2011 18:34

Picon

Re: Re[3]: [PATCH 2/2]: Hexagon atomic ops

Hi,

On 16 November 2011 10:33, Ivan Maidanski <ivmai@...> wrote:
> Hi Linas,
>
> Based on your implementation of CAS for Hexagon, I've created another CAS primitive equivalent to GCC
__sync_val_compare_and_swap one (see AO_fetch_compare_and_swap() in https://github.com/ivmai/libatomic_ops/blob/ao_val_cas/src/atomic_ops/sysdeps/gcc/hexagon.h).
> But I have no opportunity to test the code.
>
> Could you please test that code ("ao_val_cas" branch) for Hexagon?
>
> git clone git://github.com/ivmai/libatomic_ops.git -b ao_val_cas
> cd libatomic_ops; autoreconf -vif
> ./configure; make check

Looks good to me; the last few lines are:

PASS: test_stack
Performing 1000 reversals of 1000 element lists in 10 threads
Testing AO_malloc/AO_free
Succeeded
PASS: test_malloc
==================
All 4 tests passed
==================

--linas

Group

gmane.comp.programming.garbage-collection.boehmgc.

Project Web Page

Discussions about the Boehm Garbage Collector (Boehm GC).

Search Archive

Page

1 | 2

Articles in period: 16

November 2011

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane

Recent entries

[bdwgc] Convert readme to markdown. (#24) (12 May 2013)
Silently fails to allocate memory? (14 Apr 2013)
Silently fails to allocate memory? (13 Apr 2013)
Using GC in C++ via multiple inheritance (13 Apr 2013)
Silently fails to allocate memory? (12 Apr 2013)
[win32] Isn't this a race condition? (11 Apr 2013)
GC problem on Android (3 Apr 2013)
opposite of GC_exclude_static_roots (2 Apr 2013)
Mark with offset (31 Mar 2013)
Minor autoconf problem (FYI) (30 Mar 2013)
Defining USE_LIBC_PRIVATES on all glibc systems (28 Mar 2013)
32-bit on 64-bit machine (19 Mar 2013)
typed GC with marking routine... (13 Mar 2013)
Collection from unknown thread with iOS simulator (12 Mar 2013)
typed GC with marking routine... (12 Mar 2013)
Collection from unknown thread with iOS simulator (12 Mar 2013)
bogus installed gc_config_macros.h (9 Mar 2013)
Exported symbols in 7.1 missing from 7.2 (2 Mar 2013)
Possible bug: GC_realloc (15 Feb 2013)
How does Boehm GC identify values as pointers? (15 Feb 2013)
[PATCH] pthread-using tests -lpthread (13 Feb 2013)

Archives

2	May 2013
31	April 2013
28	March 2013
22	February 2013
28	January 2013
35	December 2012
67	November 2012
35	October 2012
19	September 2012
36	August 2012
68	July 2012
37	June 2012
79	May 2012
41	April 2012
89	March 2012
63	February 2012
58	January 2012
42	December 2011
16	November 2011
25	October 2011
63	September 2011
101	August 2011
17	July 2011
17	June 2011
61	May 2011
57	April 2011
60	March 2011
57	February 2011
70	January 2011
39	December 2010
34	November 2010
44	October 2010
30	September 2010
69	August 2010
84	July 2010
19	June 2010
41	May 2010
54	April 2010
45	March 2010
87	February 2010
15	January 2010
90	December 2009
69	November 2009
158	October 2009
276	September 2009
142	August 2009
44	July 2009
131	June 2009
134	May 2009
22	April 2009
16	March 2009
27	February 2009
12	January 2009
26	December 2008
73	November 2008
113	October 2008
36	September 2008
64	August 2008
18	July 2008
11	June 2008
32	May 2008
14	April 2008
19	March 2008
60	February 2008
70	January 2008
103	December 2007
15	November 2007
42	October 2007
8	September 2007
40	August 2007
45	July 2007
26	June 2007
65	May 2007
47	April 2007
28	March 2007
23	February 2007
45	January 2007
16	December 2006
52	November 2006
46	October 2006
17	September 2006
22	August 2006
23	July 2006
36	June 2006
49	May 2006
39	April 2006
51	March 2006
37	February 2006
73	January 2006
15	December 2005
32	November 2005
13	October 2005
34	September 2005
32	August 2005
51	July 2005
33	June 2005
52	May 2005
32	April 2005
25	March 2005
20	February 2005
27	January 2005
21	December 2004
80	November 2004
28	October 2004
24	September 2004
34	August 2004
38	July 2004
28	June 2004
33	May 2004
43	April 2004
36	March 2004
37	February 2004
21	January 2004
19	December 2003
74	November 2003
24	October 2003
7	September 2003
8	August 2003
28	July 2003
25	June 2003
33	May 2003
2	April 2003
19	March 2003
14	February 2003
11	January 2003
10	December 2002
1	November 2002
2	October 2002

Design

Zawodny | Right Menu | Left Menu

Your Own Design

Paste the URL of your CSS below. Download CSS template