11:20 p.m.
Raymond Toy <raymond.toy(a)ericsson.com> writes:
Built from a full tarball, but I grabbed a patch to make
src/syssignal.h be the same as the CVS version.
Everything seems to be ok, but it seems a quite a bit slower and more
consy than 21.5-b18. No hard numbers though.
Hi Ray!
M-x show-memory-usage
can easily get you the data, see the bottom section of its output.
Make sure not to disable debugging support, which also defines
MEMORY_USAGE_STATS.
Are you building with the exact same configuration you used for
21.5-b18?
Best regards,
Adrian
Ray
> XEmacs Build Report generated by emacs-version
> 21.5 (beta20) "cilantro" (+CVS-20050310) XEmacs Lucid
> with system-configuration
> sparc-sun-solaris2.8
> follows:
> Contents of /apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/Installation:
> (Output from most recent run of ./configure)
uname -a: SunOS brtps071 5.8 Generic_117350-02 sun4u sparc SUNW,Sun-Blade-1500
../configure '--with-compiler=/apps/gnu/solaris2.8/gcc-3.4.3/bin/gcc'
'--verbose' '--enable-mule' '--with-ldap=yes'
'--enable-modules=no' '--with-pop' '--enable-sound=noesd'
'--prefix=/apps/public/XEmacs/21.5'
'--bindir=/apps/public/XEmacs/21.5/bin/solaris2.8'
'--enable-dialogs=athena' '--enable-widgets=athena'
'--with-site-includes=/apps/public/solaris2.7/usr/openwin/include
/apps/public/solaris2.7/openldap/include /apps/public/solaris2.7/include'
'--with-site-libraries=/apps/public/solaris2.7/usr/openwin/lib
/apps/public/solaris2.7/openldap/lib /apps/public/solaris2.7/lib'
XEmacs 21.5-b20 "cilantro" (+CVS-20050310) configured for
`sparc-sun-solaris2.8'.
WARNING: --------------------------------------------------------------
WARNING: This was produced from a new autoconf-2.5 based configuration.
WARNING: If this configuration seems buggy then copy etc/configure-2.13
WARNING: to configure and try again. Also please report the bug.
WARNING: --------------------------------------------------------------
Compilation Environment and Installation Defaults:
Source code location: /apps/public/XEmacs/src/xemacs-21.5.20
Installation prefix: /apps/public/XEmacs/21.5
Additional header files: /apps/public/solaris2.7/usr/openwin/include
/apps/public/solaris2.7/openldap/include /apps/public/solaris2.7/include
Additional libraries: /apps/public/solaris2.7/usr/openwin/lib
/apps/public/solaris2.7/openldap/lib /apps/public/solaris2.7/lib
Runtime library search path:
/usr/ccs/lib:/apps/public/solaris2.7/usr/openwin/lib:/apps/public/solaris2.7/openldap/lib:/apps/public/solaris2.7/lib:/usr/openwin/lib:/apps/gnu/solaris2.7/gcc-3.4.3/bin/../lib/gcc/sparc-sun-solaris2.8/3.4.3/../../..:/apps/gnu/solaris2.8/gcc-3.4.3/lib/gcc/sparc-sun-solaris2.8/3.4.3/../../..
Operating system description file: `s/sol2.h'
Machine description file: `m/sparc.h'
Compiler version: gcc (GCC) 3.4.3
- GCC specs file:
/apps/gnu/solaris2.7/gcc-3.4.3/bin/../lib/gcc/sparc-sun-solaris2.8/3.4.3/specs
- Compiler command: /apps/gnu/solaris2.8/gcc-3.4.3/bin/gcc -Wall
-Wno-switch -Wundef -Wsign-compare -Wno-char-subscripts -Wpacked -Wpointer-arith
-Wunused-parameter -g
libc version: SUNWcsl 11.8.0,REV=2000.01.08.18.12
Relocating allocator for buffers: yes
GNU version of malloc: yes
Window System:
Compiling in support for the X window system:
- X Windows headers location: /usr/openwin/include
- X Windows libraries location: /usr/openwin/lib
- Handling WM_COMMAND properly.
Compiling in support for the Athena widget set:
- Athena headers location: X11/Xaw
- Athena library to link: Xaw
Using Lucid menubars.
Using Lucid scrollbars.
Using Athena dialog boxes.
Using Athena native widgets.
TTY:
Images:
Compiling in support for GIF images (builtin).
Compiling in support for XPM images.
Compiling in support for PNG images.
Compiling in support for JPEG images.
Compiling in support for TIFF images.
Sound:
Compiling in support for sound (native).
Compiling in support for NAS (network audio system).
Databases:
Compiling in support for Berkeley database.
Compiling in support for DBM.
Compiling in support for LDAP.
Internationalization:
Compiling in support for Mule (multi-lingual Emacs).
Compiling in support for XIM (X11R5+ I18N input method).
- Using Motif to provide XIM support.
Mail:
Compiling in support for POP mail retrieval.
Compiling in support for "file" mail spool file locking method.
Other Features:
Inhibiting IPv6 canonicalization at startup.
Compiling in support for ToolTalk.
Compiling in support for extra debugging code.
Compiling in support for runtime error checking.
WARNING: ---------------------------------------------------------
WARNING: XEmacs will run noticeably more slowly as a result.
WARNING: Error checking is on by default for XEmacs beta releases.
WARNING: ---------------------------------------------------------
/apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/xemacs-make-install.err not
found!
/apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/xemacs-make-check.err not found!
/apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/xemacs-make-check-temacs.err not
found!
/apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/xemacs-make-all.err not found!
/apps/public/XEmacs/src/xemacs-21.5.20/build-gcc-3.4.3/beta.err not found!
--
Adrian Aichner
mailto:adrian＠xemacs.org
http://www.xemacs.org/
10:22 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
>>>> "Adrian" == Adrian Aichner
<adrian(a)xemacs.org> writes:
Adrian> Raymond Toy <raymond.toy(a)ericsson.com> writes:
> Built from a full tarball, but I grabbed a patch to make
> src/syssignal.h be the same as the CVS version.
>
> Everything seems to be ok, but it seems a quite a bit slower and more
> consy than 21.5-b18. No hard numbers though.
Adrian> Hi Ray!
Hello!
I think I should qualify the slowness a bit. It seems that gnus is
slow. I have some folders in gnus with about 1000 unread entries. It
seems b20 takes a lot longer and gc's quite a bit when getting the
summary buffer.
Otherwise, I haven't noticed slowness anywhere else, but I don't think
my normal usage requires lots of heavy stuff to run, except gnus.
Adrian> M-x show-memory-usage
Adrian> can easily get you the data, see the bottom section of its output.
Adrian> Make sure not to disable debugging support, which also defines
Adrian> MEMORY_USAGE_STATS.
I never turn off any debugging stuff on the betas. I'll send the
numbers and stuff later.
Adrian> Are you building with the exact same configuration you used for
Adrian> 21.5-b18?
Because I never remember how to configure xemacs, I basically copy the
config (in config.status) from previous versions. However, I did
change from gcc 3.1 to gcc 3.4.3 in this build.
Would it matter if I changed back to 3.1? Would you be interested in
the results?
Thanks,
Ray
11:47 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
Raymond Toy <raymond.toy(a)ericsson.com> writes:
>>>>> "Adrian" == Adrian Aichner
<adrian(a)xemacs.org> writes:
Adrian> Raymond Toy <raymond.toy(a)ericsson.com> writes:
>> Built from a full tarball, but I grabbed a patch to make
>> src/syssignal.h be the same as the CVS version.
>>
>> Everything seems to be ok, but it seems a quite a bit slower and more
>> consy than 21.5-b18. No hard numbers though.
Adrian> Hi Ray!
Hello!
I think I should qualify the slowness a bit. It seems that gnus is
slow. I have some folders in gnus with about 1000 unread entries. It
Oh, and I feel guilty when I have 15 or 20 unread messages, because I
keep them unread to make them serve as a to-do list :-)
And you did not change gnus-version or configuration at all?
gnus-use-byte-compile could be of interest.
A gnus pre-release package just went out.
You're not using that one yet, I guess?
seems b20 takes a lot longer and gc's quite a bit when getting
the
summary buffer.
Providing profiling output for 21.5-b18 and 21.5-b20 might be useful.
Just use
M-x profile-command
or
M-x profile-key-sequence
whichever is more convenience for your usage pattern.
Might want to include output of (describe-installation) verbatim into
that report.
That should make it easy to use ediff-regions-wordwise or somesuch
between the two reports.
Otherwise, I haven't noticed slowness anywhere else, but I don't think
my normal usage requires lots of heavy stuff to run, except gnus.
Adrian> M-x show-memory-usage
Adrian> can easily get you the data, see the bottom section of its output.
Adrian> Make sure not to disable debugging support, which also defines
Adrian> MEMORY_USAGE_STATS.
I never turn off any debugging stuff on the betas. I'll send the
numbers and stuff later.
Great.
I mostly use show-memory-usage to let it guide me to kill a few big
(usually buffers starting in SPC, which aren't shown) buffers when my
XEmacs gets too big.
Exiting and starting gnus also helps quite a bit.
Adrian> Are you building with the exact same configuration you used for
Adrian> 21.5-b18?
Because I never remember how to configure xemacs, I basically copy the
config (in config.status) from previous versions. However, I did
change from gcc 3.1 to gcc 3.4.3 in this build.
Why not save the configure command line and rerun that?
Is it too late for that now?
My build XEmacs Package features build-settings[1] since a little while,
which may help.
Would it matter if I changed back to 3.1? Would you be interested in
the results?
I wouldn't personally use that data.
Others might.
Unless you want to switch back gcc versions for other reasons, I would
rather have you provide profiling output and configuration
information.
Narrowing in on the fat rabbit would help.
Best regards,
Adrian
Thanks,
Ray
Footnotes:
[1] http://www.xemacs.org/Documentation/packages/html/build_4.html#SEC6
--
Adrian Aichner
mailto:adrian＠xemacs.org
http://www.xemacs.org/
3:05 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
>>>> "Adrian" == Adrian Aichner
<adrian(a)xemacs.org> writes:
Adrian> Raymond Toy <raymond.toy(a)ericsson.com> writes:
> I think I should qualify the slowness a bit. It seems that gnus
is
> slow. I have some folders in gnus with about 1000 unread entries. It
Adrian> Oh, and I feel guilty when I have 15 or 20 unread
Adrian> messages, because I keep them unread to make them serve as
Adrian> a to-do list :-)
I'm too lazy to sort them out even though I should. They're just old
messages with useful info that I want to remember. :-) When I had a
slow machine, gnus was too slow after about 500 or so messages, but I
have a faster machine now. :-)
Adrian> And you did not change gnus-version or configuration at all?
Oops. I also downloaded the 2005-03-07 sumo packages. I guess I
should try b18 with these to see how that is.
Adrian> gnus-use-byte-compile could be of interest.
Adrian> A gnus pre-release package just went out.
Adrian> You're not using that one yet, I guess?
No. I used to try to be on the bleeding edge, but I've run out of
blood.
Adrian> I mostly use show-memory-usage to let it guide me to kill a few big
Adrian> (usually buffers starting in SPC, which aren't shown) buffers when my
Adrian> XEmacs gets too big.
Adrian> Exiting and starting gnus also helps quite a bit.
>
Adrian> Are you building with the exact same
configuration you
Adrian> used for
Adrian> 21.5-b18?
>
>> Because I never remember how to configure
xemacs, I basically copy the
>> config (in config.status) from previous versions. However, I did
>> change from gcc 3.1 to gcc 3.4.3 in this build.
Adrian> Why not save the configure command line and rerun that?
"head config.status" gives the configure command I used. Useful when
experimenting with different configs, provided I remember to do the
configs in different directories.
Adrian> Unless you want to switch back gcc versions for other
Adrian> reasons, I would rather have you provide profiling output
Adrian> and configuration information.
The version doesn't really matter. I just decided to try 3.4.3 since
it was the latest around. I'm pretty sure some versions after 3.1
would produce a broken xemacs, so I stayed with 3.1, until now.
More info coming soon,
Ray
3:09 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
Raymond Toy <raymond.toy(a)ericsson.com> wrote:
"head config.status" gives the configure command I used.
Useful when
experimenting with different configs, provided I remember to do the
configs in different directories.
I am used to that idiom, also. Unfortunately, autoconf 2.59 has put an
end to it. The configuration information is now buried far down in
config.status, and with lots of extra quote marks to boot. It's too bad
the autoconf team did that. Having the config information at the top
was very convenient.
--
Jerry James
http://www.ittc.ku.edu/~james/
4:32 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
>>>> "Jerry" == Jerry James
<james(a)xemacs.org> writes:
Jerry> Raymond Toy <raymond.toy(a)ericsson.com> wrote:
> "head config.status" gives the configure command I
used. Useful when
> experimenting with different configs, provided I remember to do the
> configs in different directories.
Jerry> I am used to that idiom, also. Unfortunately, autoconf 2.59 has put an
Jerry> end to it. The configuration information is now buried far down in
Jerry> config.status, and with lots of extra quote marks to boot. It's too
bad
Jerry> the autoconf team did that. Having the config information at the top
Jerry> was very convenient.
You're right! I didn't check this version. But "config.status
--version" works well enough now. Requires a little more cutting and
pasting, though.
Ray
3:19 a.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
>>>> "Raymond" == Raymond Toy
<raymond.toy(a)ericsson.com> writes:
>>>> "Adrian" == Adrian Aichner
<adrian(a)xemacs.org> writes:
Adrian> Raymond Toy <raymond.toy(a)ericsson.com> writes:
> Built from a full tarball, but I grabbed a patch to make
> src/syssignal.h be the same as the CVS version.
>
> Everything seems to be ok, but it seems a quite a bit slower and more
> consy than 21.5-b18. No hard numbers though.
Adrian> Hi Ray!
Raymond> I think I should qualify the slowness a bit. It seems
Raymond> that gnus is slow. I have some folders in gnus with
Raymond> about 1000 unread entries. It seems b20 takes a lot
Raymond> longer and gc's quite a bit when getting the summary
Raymond> buffer.
I think there have been changes to both the regexp code and to
font-lock which might affect Gnus.
Are you running with error-checking on in Mule? This increases
computation by a full factor of N (N = buffer size) to many
operations, as with a variable-width representation of characters
checking position-computing code may require starting from the
beginning....
(setq gc-cons-threshold (* 8 1000 1000)) is probably a good idea on
most hardware; is there anybody trying to run XEmacs in 16MB anymore?
Or you could try (setq gc-cons-percentage 40). (Aidan proposed a
patch to make gc-cons-threshold 8MB by default, but Ben basically said
"I'm getting rid of gc-cons-threshold in favor of gc-cons-percentage,
which gives constant amortized GC-ing", so nothing happened. But I
don't know if gc-cons-percentage is fully implemented in 21.5.20 or not.)
Raymond> Would it matter if I changed back to 3.1? Would you be
Raymond> interested in the results?
I don't think it's worth it.
The only optimization I've heard of that actually produces measurable
speedup is -Os ;-). You might try that.
--
School of Systems and Information Engineering http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
3:10 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
>>>> "Stephen" == Stephen J Turnbull
<stephen(a)xemacs.org> writes:
>>>> "Raymond" == Raymond Toy
<raymond.toy(a)ericsson.com> writes:
>>>> "Adrian" == Adrian Aichner
<adrian(a)xemacs.org> writes:
Stephen> I think there have been changes to both the regexp code and to
Stephen> font-lock which might affect Gnus.
Stephen> Are you running with error-checking on in Mule? This increases
Stephen> computation by a full factor of N (N = buffer size) to many
Stephen> operations, as with a variable-width representation of characters
Stephen> checking position-computing code may require starting from the
Stephen> beginning....
I've always run with Mule enabled with error-checking. I didn't know
it made it that bad, otherwise I probably would have quit building
with Mule since I'm illiterate anyway.
Stephen> (setq gc-cons-threshold (* 8 1000 1000)) is probably a good idea on
Stephen> most hardware; is there anybody trying to run XEmacs in 16MB anymore?
Heh. My gc-cons-threshold is 500000. I don't think I set that
myself.
I'll try it out.
Stephen> The only optimization I've heard of that actually produces measurable
Stephen> speedup is -Os ;-). You might try that.
From what you say, killing Mule would be a bigger win.
Thanks for your help,
Ray
12:18 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
Are you running with error-checking on in Mule? This increases
BTW, what do you call error-checking? There are several flags here:
OPTIMIZED_BUILD
DEBUG_XEMACS
ERROR_CHECK_ALL
I assume it is ERROR_CHECK_ALL you were talking about.
Best,
Fabrice
6:28 p.m.
New subject: [Ok, with patch] XEmacs 21.5-b20 "cilantro" (+CVS-20050310) s parc-sun-solaris2.8
Fabrice Popineau <Fabrice.Popineau(a)supelec.fr> writes:
> Are you running with error-checking on in Mule? This increases
BTW, what do you call error-checking? There are several flags here:
OPTIMIZED_BUILD
DEBUG_XEMACS
ERROR_CHECK_ALL
I assume it is ERROR_CHECK_ALL you were talking about.
Yep, that should be the one.
Best,
Fabrice
--
Adrian Aichner
mailto:adrian＠xemacs.org
http://www.xemacs.org/
6978
days inactive
6981
days old
9 comments
5 participants
participants (5)
-
Adrian Aichner -
Fabrice Popineau -
Jerry James -
Raymond Toy -
Stephen J. Turnbull