Emacs costs 100M memory on Windows XP

Emacs costs 100M memory on Windows XP

[Zhang Wei]

On Windows XP, with font backend enabled, Emacs costs 100M or more
memory, while it just cost 20M or so before, why does it consume so
much merory? 4 times than normal configuration.

In GNU Emacs 23.0.60.1 (i386-mingw-nt5.1.2600)
 of 2008-02-02 on BREPNB
Windowing system distributor `Microsoft Corp.', version 5.1.2600
configured using `configure --with-gcc (3.4) --cflags -DUSE_FONT_BACKEND'

Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: chs
  value of $XMODIFIERS: nil
  locale-coding-system: cp936
  default-enable-multibyte-characters: t

Major mode: Lisp Interaction

Minor modes in effect:
  tooltip-mode: t
  mouse-wheel-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  font-lock-mode: t
  blink-cursor-mode: t
  global-auto-composition-mode: t
  auto-composition-mode: t
  auto-compression-mode: t
  line-number-mode: t
  transient-mark-mode: t

Re: Emacs costs 100M memory on Windows XP

[Eli Zaretskii]

> Date: Sat, 2 Feb 2008 19:02:58 +0800
> From: "Zhang Wei" <id.brep@gmail.com>
> 
> On Windows XP, with font backend enabled, Emacs costs 100M or more
> memory

How did you work around the problem with charprop.el and other files
generated by admin/unidata-gen.el?  Did you generate those files
manually, outside "make bootstrap"?

Re: Emacs costs 100M memory on Windows XP

[Eric Lilja]

Eli Zaretskii wrote:
>> Date: Sat, 2 Feb 2008 19:02:58 +0800
>> From: "Zhang Wei" <id.brep@gmail.com>
>>
>> On Windows XP, with font backend enabled, Emacs costs 100M or more
>> memory
> 
> How did you work around the problem with charprop.el and other files
> generated by admin/unidata-gen.el?  Did you generate those files
> manually, outside "make bootstrap"?
> 
> 
> 

I'm also unable to build on windows with --enable-font-backend (haven't 
tried without yet), it fails because No rule to make target 
`../lisp/international/charprop.el'. I'm also wondering what 
--enable-font-backend does on windows...

Furthermore, I'm getting loads of these warnings:
Warning: Default coding system `iso-latin-1' disagrees with
system codeset `cp1252' for this locale.
For most if not all lisp files. Did I do something wrong in the 
configure step for this warning to appear?

- Eric Lilja

Re: Emacs costs 100M memory on Windows XP

[Eli Zaretskii]

> From: Eric Lilja <mindcooler@gmail.com>
> Date: Sat, 02 Feb 2008 13:31:23 +0100
> 
> Eli Zaretskii wrote:
> >> Date: Sat, 2 Feb 2008 19:02:58 +0800
> >> From: "Zhang Wei" <id.brep@gmail.com>
> >>
> >> On Windows XP, with font backend enabled, Emacs costs 100M or more
> >> memory
> > 
> > How did you work around the problem with charprop.el and other files
> > generated by admin/unidata-gen.el?  Did you generate those files
> > manually, outside "make bootstrap"?
> > 
> > 
> > 
> 
> I'm also unable to build on windows with --enable-font-backend (haven't 
> tried without yet), it fails because No rule to make target 
> `../lisp/international/charprop.el'.

This happens because the Windows build does not run
admin/unidata/Makefile.  I'm working on fixing that.

> I'm also wondering what --enable-font-backend does on windows...

Same here.

Re: Emacs costs 100M memory on Windows XP

[Zhang Wei]

On 2/2/08, Eli Zaretskii <eliz@gnu.org> wrote:
> > Date: Sat, 2 Feb 2008 19:02:58 +0800
> > From: "Zhang Wei" <id.brep@gmail.com>
> >
> > On Windows XP, with font backend enabled, Emacs costs 100M or more
> > memory
>
> How did you work around the problem with charprop.el and other files
> generated by admin/unidata-gen.el?  Did you generate those files
> manually, outside "make bootstrap"?
>

I just copy them from the emacs-unicode-2 branch.

Re: Emacs costs 100M memory on Windows XP

[Jason Rumney]

Eli Zaretskii wrote:
> This happens because the Windows build does not run
> admin/unidata/Makefile.  I'm working on fixing that.
>   
I added some rules to build charprop.el yesterday. However, the DOC file 
is built on Windows BEFORE temacs, so we cannot generate charprop.el in 
time for the build to succeed.

Re: Emacs costs 100M memory on Windows XP

[Jason Rumney]

Eric Lilja wrote:
> I'm also wondering what --enable-font-backend does on windows...

It does the same as other platforms. It enables the new font backend 
which better handles unicode fonts.

Re: Emacs costs 100M memory on Windows XP

[Eli Zaretskii]

> Date: Sat, 02 Feb 2008 13:10:39 +0000
> From: Jason Rumney <jasonr@gnu.org>
> CC: Eric Lilja <mindcooler@gmail.com>, emacs-devel@gnu.org
> 
> Eli Zaretskii wrote:
> > This happens because the Windows build does not run
> > admin/unidata/Makefile.  I'm working on fixing that.
> >   
> I added some rules to build charprop.el yesterday. However, the DOC file 
> is built on Windows BEFORE temacs, so we cannot generate charprop.el in 
> time for the build to succeed.

I think you added those rules in the wrong place, or maybe they are
not enough to bootstrap.  DOC is indeed built before temacs, but we
already have built the bootstrap version of temacs and emacs before
that, so we could use those bootstrap versions to produce charprop.el
(and other files generated from UnicodeData.txt).  This is what the
Unix build process does.

I added the necessary rules along the above ideas, and I'm
bootstrapping now.  If the bootstrap runs to completion, I will commit
the changes I have so far.

Re: Emacs costs 100M memory on Windows XP

[Eli Zaretskii]

> Date: Sat, 02 Feb 2008 13:14:17 +0000
> From: Jason Rumney <jasonr@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> Eric Lilja wrote:
> > I'm also wondering what --enable-font-backend does on windows...
> 
> It does the same as other platforms. It enables the new font backend 
> which better handles unicode fonts.

Do we need to fix the build without it (which currently fails on
Windows), or should we make --enable-font-backend the default?

Re: Emacs costs 100M memory on Windows XP

[Eli Zaretskii]

> Date: Sat, 2 Feb 2008 19:02:58 +0800
> From: "Zhang Wei" <id.brep@gmail.com>
> 
> On Windows XP, with font backend enabled, Emacs costs 100M or more
> memory

It's "only" 87MB for me, but that is also way too much.

Re: Emacs costs 100M memory on Windows XP

[Jason Rumney]

Eli Zaretskii wrote:
>> Date: Sat, 02 Feb 2008 13:14:17 +0000
>> From: Jason Rumney <jasonr@gnu.org>
>> Cc: emacs-devel@gnu.org
>>
>> Eric Lilja wrote:
>>     
>>> I'm also wondering what --enable-font-backend does on windows...
>>>       
>> It does the same as other platforms. It enables the new font backend 
>> which better handles unicode fonts.
>>     
>
> Do we need to fix the build without it (which currently fails on
> Windows), or should we make --enable-font-backend the default?
>   

I think the original plan was to make the new font backend the default 
after the initial flood of post-merge bug reports died down, but if the 
old font code is going to be a major source of such reports, maybe it is 
better to make it the default sooner.

Re: Emacs costs 100M memory on Windows XP

[Kenichi Handa]

In article <47A4DAED.4000801@gnu.org>, Jason Rumney <jasonr@gnu.org> writes:

> > Do we need to fix the build without it (which currently fails on
> > Windows), or should we make --enable-font-backend the default?
> >   

> I think the original plan was to make the new font backend the default 
> after the initial flood of post-merge bug reports died down, but if the 
> old font code is going to be a major source of such reports, maybe it is 
> better to make it the default sooner.

I agree.

And I also had the plan of deleting regacy font handling
codes as soon as it is supported also on non-Unix
non-GNU-Linux system (i.e. Windows and Mac).  When Emacs.app
(Mac Cocoa port) is merged into the trunk, I'd like to start
working on it.

---
Kenichi Handa
handa@ni.aist.go.jp

Re: Emacs costs 100M memory on Windows XP

[Jason Rumney]

Kenichi Handa wrote:
>> I think the original plan was to make the new font backend the default 
>> after the initial flood of post-merge bug reports died down, but if the 
>> old font code is going to be a major source of such reports, maybe it is 
>> better to make it the default sooner.
>>     
>
> I agree.
>   

The problems on Windows have been fixed, so there is no urgency now.

Re: Emacs costs 100M memory on Windows XP

[Juanma Barranquero]

On Feb 4, 2008 1:50 PM, Jason Rumney <jasonr@gnu.org> wrote:

> The problems on Windows have been fixed, so there is no urgency now.

And the new font backend seems significantly slower.

             Juanma

Re: Emacs costs 100M memory on Windows XP

[Miles Bader]

"Juanma Barranquero" <lekktu@gmail.com> writes:
>> The problems on Windows have been fixed, so there is no urgency now.
>
> And the new font backend seems significantly slower.

In what case?

I didn't notice any obviousy slowdown when switching to the new font
backend, even on my home machine, which is fairly slow (450 MHz P3).

-Miles

-- 
People who are more than casually interested in computers should have at
least some idea of what the underlying hardware is like.  Otherwise the
programs they write will be pretty weird.  -- Donald Knuth

Re: Emacs costs 100M memory on Windows XP

[Juanma Barranquero]

> I didn't notice any obviousy slowdown when switching to the new font
> backend, even on my home machine, which is fairly slow (450 MHz P3).

It is perhaps only an issue of the Windows port.

I can see the problem, for example, by visiting any file and just
scrolling down. With the new backend, when the cursor reaches the
bottom and the window is recentered, the delay in redrawing the bottom
half is clearly noticeable (often half a second or more). That's on a
3GHz Pentium IV with Windows XP Pro. With the old backend, the redraw
is instantaneous.

             Juanma

Re: Emacs costs 100M memory on Windows XP

[Jason Rumney]

Eli Zaretskii wrote:
>> Date: Sat, 2 Feb 2008 19:02:58 +0800
>> From: "Zhang Wei" <id.brep@gmail.com>
>>
>> On Windows XP, with font backend enabled, Emacs costs 100M or more
>> memory
>>     
>
> It's "only" 87MB for me, but that is also way too much.
>   

The bug should be fixed now, the minimum width and height of characters 
was being initialised to 1 (pixel), and not updated to the real value by 
the new font code, so redisplay was allocating far more memory than it 
needed to keep track of the displayed characters.

Re: Emacs costs 100M memory on Windows XP

[Zhang Wei]

On 2/6/08, Jason Rumney <jasonr@gnu.org> wrote:
> The bug should be fixed now, the minimum width and height of characters
> was being initialised to 1 (pixel), and not updated to the real value by
> the new font code, so redisplay was allocating far more memory than it
> needed to keep track of the displayed characters.

The bug is fixed, thank you.

Re: Emacs costs 100M memory on Windows XP

[YAMAMOTO Mitsuharu]

>>>>> On Mon, 04 Feb 2008 21:33:18 +0900, Kenichi Handa <handa@ni.aist.go.jp> said:

>> I think the original plan was to make the new font backend the
>> default after the initial flood of post-merge bug reports died
>> down, but if the old font code is going to be a major source of
>> such reports, maybe it is better to make it the default sooner.

> I agree.

> And I also had the plan of deleting regacy font handling codes as
> soon as it is supported also on non-Unix non-GNU-Linux system
> (i.e. Windows and Mac).  When Emacs.app (Mac Cocoa port) is merged
> into the trunk, I'd like to start working on it.

The Mac Carbon port in the trunk already does not compile regardless
of its font backend setting.  Also, the Mac Cocoa port does not
support the legacy font code now.  That says, switching to the new
font backend does not do harm to either of Mac ports with respect to
the current status.  So I think it's not necessary to wait for the
merger of the Mac Cocoa port to switch to the new font backend.

				     YAMAMOTO Mitsuharu
				mituharu@math.s.chiba-u.ac.jp