[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gcl-devel] Re: GCL 2.6.6 is released
|
From: |
Camm Maguire |
|
Subject: |
Re: [Gcl-devel] Re: GCL 2.6.6 is released |
|
Date: |
23 Jan 2005 14:55:19 -0500 |
|
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
Greetings!
"Matt Kaufmann" <address@hidden> writes:
> Thanks! I've succesffully built ACL2 2.9.1 on GCL 2.6.6 at AMD, following
> your
> GCL configure suggestion.
>
> What's more, I've successfully run that ACL2/GCL image on the project at AMD
> on
> which I'm currently working. As you know, GCL (2.6.5) has consistently run
> substantially faster than Allegro CL 6.2 on linux on the ACL2 regression suite
> (see "Performance Comparisons" on
> http://www.cs.utexas.edu/users/moore/acl2/v2-9/new.html). However, I'm
> finding
Great!
> GCL somewhat slower than Allegro on my current AMD project. There are two
> separate calls of make in each case, each invoking ACL2 many times (something
> like 10 times for the first part and 70 times for the second part). All runs
> are on linux.
>
> Allegro CL 6.2 on first part:
> 315.060u 7.320s 5:28.07 98.2% 0+0k 0+0io 114490pf+0w
>
> GCL 2.6.6 on first part:
> 381.680u 6.640s 6:33.74 98.6% 0+0k 0+0io 118537pf+0w
>
> Allegro CL 6.2 on second part:
> 2150.730u 34.330s 37:02.11 98.3% 0+0k 0+0io 866079pf+0w
>
> GCL 2.6.6 on second part:
> 2747.590u 28.160s 46:50.37 98.7% 0+0k 0+0io 786285pf+0w
>
> I spent some time browsing the logs and found the following as some of the
> largest time discrepancies in absolute (not percentage) terms.
>
> 1:34 gcl vs. 1:17 allegro
> 1:04 gcl vs. 0:53 allegro
> 2:37 gcl vs. 2:31 allegro
> 3:07 gcl vs. 3:03 allegro
> 4:27 gcl vs. 4:00 allegro
> 3:40 gcl vs. 2:35 allegro
>
Well, somewhat disappointing -- but these are always opportunities, right?
> As usual, I can't send you any of the files (sigh) as they involve AMD stuff.
> But I've made a note of which files correspond to the six comparisons above.
> I guess the most promising one to look at is the last of these six. If you
> send me (or point me to) instructions for how to profile or otherwise get
> information for GCL 2.6.6, I'd be happy to push the appropriate button and
> send
> you what I can.
>
For profiling, rebuild gcl with --enable-gprof, and then rebuild acl2
on top of this. To profile a command, sandwich it between
(si::gprof-start) and (si::gprof-quit). The profiling overhead itself
will be noticeable, but reported times should retain their
proportionality in a non-profiling image.
The first step is always to separate out the gc component of
comparable gcl and allegro runs. After that, if undesirable
performance persists, I will most likely have to see the C code
produced from a representative lisp snippet. I'm assuming you are
proclaiming everything the way you do in stock acl2, which is
exemplary in this regard.
And of course if you can isolate this enough to identify an analogous
run on stock acl2 where the performance diff is the other way around,
that would be very helpful too.
As you may know, I'm mostly away from email and office until 2/15, so
in depth replies will likely have to wait until then.
Take care,
> -- Matt
> X-Server-Uuid: 5FC0E2DF-CD44-48CD-883A-0ED95B391E89
> cc: address@hidden
> From: "Camm Maguire" <address@hidden>
> Date: 21 Jan 2005 19:06:02 -0500
> X-MMS-Spam-Filter-ID: A2005012109_IBF_2.0.0
> X-WSS-ID: 6DEF47612CW3800652-01-01
> X-Spam-Status: No, hits=0 required=5 tests=
> X-Spam-Level:
> X-Scanned-By: MIMEDefang 2.39
>
> Greetings!
>
> "Matt Kaufmann" <address@hidden> writes:
>
> > Thanks, Camm. I unpacked the .deb (but not as root, so usr/ and etc/
> were put
> > under the current directory, and I edited usr/bin/gcl accordingly). But
> I
> > guess we don't have the right libraries. Is there an easy workaround?
> (I
> > don't think I want to ask the sysadmins to mess with the libraries;
> things need
> > to be kept pretty stable around here.)
> >
> > rainer:/u/acl2/gcl[ 600 ] --> /u/acl2/gcl/gcl-2.6.6/usr/bin/gcl
> > /u/acl2/gcl/gcl-2.6.6/usr/lib/gcl-2.6.6/unixport/saved_gcl:
> /lib/libc.so.6: version `GLIBC_2.3' not found (required by
> /u/acl2/gcl/gcl-2.6.6/usr/lib/gcl-2.6.6/unixport/saved_gcl)
> > rainer:/u/acl2/gcl[ 601 ] -->
> /u/acl2/gcl/gcl-2.6.6/usr/lib/gcl-2.6.6/unixport/saved_gcl
> > /u/acl2/gcl/gcl-2.6.6/usr/lib/gcl-2.6.6/unixport/saved_gcl:
> /lib/libc.so.6: version `GLIBC_2.3' not found (required by
> /u/acl2/gcl/gcl-2.6.6/usr/lib/gcl-2.6.6/unixport/saved_gcl)
> > rainer:/u/acl2/gcl[ 602 ] -->
> >
>
> Alas, no, no workaround here.
>
> > Here is the tail of the log when I tried, earlier, to build gcl myself:
> >
> > gcc -c -Wall -DVOL=volatile -fsigned-char -pipe -O3 -fomit-frame-pointer
> -I/u/acl2/gcl/gcl-2.6.6/o -I../h -I../gcl-tk fat_string.c
> > In file included from /usr/include/bfd.h:103,
> > from fat_string.c:17:
> >
> /tool/pandora/.package/gcc-2.95.3/lib/gcc-lib/i686-pc-linux-gnu/2.95.3/include/stdbool.h:11:
> conflicting types for `bool'
> > ../h/../h/object.h:68: previous declaration of `bool'
> > make[1]: *** [fat_string.o] Error 1
> > make[1]: Leaving directory `/u/acl2/gcl/gcl-2.6.6/o'
> > make: *** [unixport/saved_pre_gcl] Error 2
> >
>
> This is simple -- ./configure --disable-statsysbfd --enable-locbfd &&
> make
>
> Please let me know if problems persist.
>
> Take care,
>
>
> > Thanks --
> > -- Matt
> > cc: address@hidden
> > From: "Camm Maguire" <address@hidden>
> > Date: 20 Jan 2005 22:44:42 -0500
> > User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
> > X-WSS-ID: 6DEEA55D3749131-01-01
> > Content-Type: text/plain;
> > charset=us-ascii
> > X-Spam-Status: No, hits=0 required=5 tests=
> > X-Spam-Level:
> > X-Scanned-By: MIMEDefang 2.39
> >
> > Greetings! You can retrieve a .deb from
> > http://packages.debian.org/unstable/interpreters/gcl and follow the
> > HOWTO-UNPACK-DEBS, which is basically to ar x the .deb and then tar
> > zxf the data.tar.gz.
> >
> > Please let me know if you have difficulties. Likewise if you want to
> > forward your failed build log.
> >
> > Take care,
> >
> > "Matt Kaufmann" <address@hidden> writes:
> >
> > > Hi, Camm --
> > >
> > > It turns out that for some reason I can't see ftp.gnu.org from
> work, but I can
> > > see if from home. So I've fetched the tar.gz file.
> > >
> > > But I'm getting an error during the make, probably because we have
> an old
> > > version of gcc by default. Rather than get that working, I'd be
> delighted to
> > > download a binary x86 distribution of GCl 2.6.6, but I couldn't
> find one (we're
> > > running Redhat rather than Debian, but I doubt that this matters).
> Is one
> > > available?
> > >
> > > Thanks --
> > > -- Matt
> > > Resent-From: address@hidden
> > > Resent-To: address@hidden
> > > cc: address@hidden
> > > From: "Camm Maguire" <address@hidden>
> > > Date: 19 Jan 2005 19:56:09 -0500
> > > User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
> > > X-SpamAssassin-Status: No, hits=-2.6 required=5.0
> > > X-UTCS-Spam-Status: No, hits=-232 required=180
> > > X-WSS-ID: 6DF1D25D3453388-01-01
> > > Content-Type: text/plain;
> > > charset=us-ascii
> > > X-Spam-Status: No, hits=0 required=5 tests=
> > > X-Spam-Level:
> > > X-Scanned-By: MIMEDefang 2.39
> > >
> > > Greetings! That's odd, the links seem to work for me. Are you
> > > looking at http://www.gnu.org/software/gcl?
> > >
> > > In any case,
> > >
> > > You can get 2.6.6 either from
> > >
> > > ftp://ftp.gnu.org/pub/gnu/gcl/gcl-2.6.6.tar.gz
> > >
> > > or via cvs
> > >
> > > export CVS_RSH=ssh
> > > export CVSROOT=:ext:address@hidden:/cvsroot/gcl
> > > cvs -z9 -q co -d gcl-2.6.6 -r Version_2_6_6 gcl
> > >
> > > Please let me know if problems persist.
> > >
> > > Take care,
> > >
> > > Matt Kaufmann <address@hidden> writes:
> > >
> > > > Hi, Camm --
> > > >
> > > > Congratulations on the GCL 2.6.6 release! I'm interested in
> trying it out at
> > > > AMD, since you've put in that reader extension patch for
> foo::(bar foobar).
> > > >
> > > > Unfortunately, the "Download" and "Get Latest Release" links
> appear to be
> > > > broken. Any suggestions?
> > > >
> > > > Thanks --
> > > > -- Matt
> > > > X-IronPort-MID: 476076990
> > > > X-SBRS: 1.9
> > > > X-BrightmailFiltered: true
> > > > X-Ironport-AV: i="3.88,135,1102312800";
> > > > d="scan'208"; a="476076990:sNHT13365576"
> > > > User-Agent: SEMI/1.14.3 (Ushinoya) FLIM/1.14.3
> (=?ISO-8859-4?Q?Unebigory?=
> > > > =?ISO-8859-4?Q?=F2mae?=) APEL/10.3 Emacs/21.2
> (i386-debian-linux-gnu)
> > > > MULE/5.0 (SAKAKI)
> > > > Content-Type: text/plain; charset=US-ASCII
> > > > From: Camm Maguire <address@hidden>
> > > > Date: Tue, 18 Jan 2005 23:01:26 -0500
> > > > Reply-To: address@hidden
> > > > Sender: address@hidden
> > > > X-Listprocessor-Version: 8.2.10/020311/17:52 --
> ListProc(tm) by CREN
> > > > X-SpamAssassin-Status: No, hits=-2.6 required=5.0
> > > > X-UTCS-Spam-Status: No, hits=-272 required=180
> > > >
> > > > Greetings! The GCL team is happy to announce the release
> of version
> > > > 2.6.6, the latest achievement in the 'stable' (as opposed to
> > > > 'development') series. Please see
> http://www.gnu.org/software/gcl for
> > > > downloading information.
> > > >
> > > > This release is a minor modification of 2.6.5,
> > > > incorporating the known bug fixes that had accumulated on
> the GCL
> > > > errata page together with a few other fixes required for
> building axiom
> > > > on Windows. From the changelog:
> > > >
> > > > gcl (2.6.6-1) unstable; urgency=high
> > > >
> > > > * New upstream release
> > > > * Allow .data section to be first in executable, as on
> solaris. Also
> > > > allow for new bfd section size semantics
> > > > * Don't try to write map file when not using GNU ld.
> Also allow
> > > > compile-file to process pathnames with whitespace on
> Windows
> > > > * Fix corner case fixnum arithmetic on 64bit machines
> > > > * Rework gmp_wrappers semantics for older gcc
> > > > * Explicitly mprotect loaded code pages PROT_EXEC on x86
> Linux, as FC3
> > > > now requires it.
> > > > * lisp-implementation-version is GCL
> > > > * Reader extension patch allowing for foo::(bar foobar)
> semantics
> > > > * a shell script variable fix in "unixport/makefile" for
> MSYS
> > > > * __MINGW32__ malloc initialization fix in "o/alloc.c"
> > > > * Windows file/directory fixes in "o/unixfsys.c"
> > > > * MinGW32 -march in configure - removes deprecation
> warnings
> > > > * MinGW32 directory fix - "o/mingfile.c".
> > > > * Allow for sysconf to determine clock granularity at
> compile time to
> > > > fix time errors on the Itanium
> > > > * Disable SGC on macosx until the sgc/save problem can be
> fixed.
> > > > * Fix fixnum print bug on 64bit
> > > > * Fix nil types in room report
> > > > * 64bit fixes to fixnum_add and fixnum_sub
> > > > * Fix Mac SGC/save bug, at least in part
> > > >
> > > > For the latest information, capabilities, and status of GCL
> stable,
> > > > please see
> http://www.gnu.org/software/gcl/RELEASE-2.6.2.html.
> > > >
> > > > --
> > > > Camm Maguire
> address@hidden
> > > >
> ==========================================================================
> > > > "The earth is but one country, and mankind its citizens."
> -- Baha'u'llah
> > > >
> > > >
> > > >
> > >
> > > --
> > > Camm Maguire
> address@hidden
> > >
> ==========================================================================
> > > "The earth is but one country, and mankind its citizens." --
> Baha'u'llah
> > >
> > >
> > >
> > >
> > >
> >
> > --
> > Camm Maguire address@hidden
> >
> ==========================================================================
> > "The earth is but one country, and mankind its citizens." --
> Baha'u'llah
> >
> >
> >
> > _______________________________________________
> > Gcl-devel mailing list
> > address@hidden
> > http://lists.gnu.org/mailman/listinfo/gcl-devel
> >
> >
> >
>
> --
> Camm Maguire address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens." -- Baha'u'llah
>
>
>
>
--
Camm Maguire address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah