Emacs 21/X11 generating unbelieveable network traffic

Emacs 21/X11 generating unbelieveable network traffic

[David Abrahams]

As you can see from
http://groups.google.com/groups?dq=&hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=u4rc
734es.fsf%40siemens.com,
I've been trying to figure out why starting Emacs 21 on a remote machine
seems so painful. It is  generating reams of network traffic that apparently
has nothing to do with graphics. I'm beginning to think that from the
responses I've received, I'm asking in the wrong place. I also think this is
a bug which is why it's cross-posted to the emacs bugs list. If anyone knows
why this is happening, and especially how to stop it, I'd be indebted.

Thanks,
Dave

--
-----------------------------------------------------------
           David Abrahams * Boost Consulting
dave@boost-consulting.com * http://www.boost-consulting.com

Re: Emacs 21/X11 generating unbelieveable network traffic

[David Abrahams]

No Interest from developers? Someone sent me a link to the explanation,
eventually.

http://mail.gnu.org/pipermail/emacs-devel/2002-September/012633.html

Still, this is something emacs developers should want to fix, it seems to
me!

Hello? Anybody?

--
-----------------------------------------------------------
           David Abrahams * Boost Consulting
dave@boost-consulting.com * http://www.boost-consulting.com


"David Abrahams" <david.abrahams@rcn.com> wrote in message
news:anago0$ftk$1@main.gmane.org...
> As you can see from
>
http://groups.google.com/groups?dq=&hl=en&lr=&ie=UTF-8&oe=UTF-8&threadm=u4rc
> 734es.fsf%40siemens.com,
> I've been trying to figure out why starting Emacs 21 on a remote machine
> seems so painful. It is  generating reams of network traffic that
apparently
> has nothing to do with graphics. I'm beginning to think that from the
> responses I've received, I'm asking in the wrong place. I also think this
is
> a bug which is why it's cross-posted to the emacs bugs list. If anyone
knows
> why this is happening, and especially how to stop it, I'd be indebted.
>
> Thanks,
> Dave
>
> --
> -----------------------------------------------------------
>            David Abrahams * Boost Consulting
> dave@boost-consulting.com * http://www.boost-consulting.com

Re: Emacs 21/X11 generating unbelieveable network traffic

[Richard Stallman]

    Still, this is something emacs developers should want to fix, it seems to
    me!

Probably so, but we are all volunteers and only a few of us
know a lot about X.  It would be very good if someone who knows
enough about X would come and offer to help us.

Re: Emacs 21/X11 generating unbelieveable network traffic

[David Abrahams]

From: "Richard Stallman" <rms@gnu.org>


>     Still, this is something emacs developers should want to fix, it
seems to
>     me!
>
> Probably so, but we are all volunteers and only a few of us
> know a lot about X.

I know how it is. I already invested a lot just to raise a plausible
diagnosis.

> It would be very good if someone who knows
> enough about X would come and offer to help us.

I'm afraid I'm not that guy. The hints in the link I posted sound like they
should be enough to lead anyone with an interest to find the appropriate
resources.

-----------------------------------------------------------
           David Abrahams * Boost Consulting
dave@boost-consulting.com * http://www.boost-consulting.com

Re: Emacs 21/X11 generating unbelieveable network traffic

[Joe Buehler]

David Abrahams wrote:

> I've been trying to figure out why starting Emacs 21 on a remote machine
> seems so painful. It is  generating reams of network traffic that apparently
> has nothing to do with graphics. I'm beginning to think that from the

I upgraded our internal machines from 20.7 to 21.2 recently.  Most of our
development machines are on the other end of a 6 mbit link, and boy is
emacs 21.2 slooooooow.  A couple developers abandoned it for the older version.

It works fine across our local 100 mbit network, so it does look to me like
21.2 is a network hog.

Joe Buehler

Re: Emacs 21/X11 generating unbelieveable network traffic

[Ami Fischman]

Though it doesn't do any color/font optimization (the substance of the rms
post you reference in a followup article), lbxproxy does a pretty good job
of unifying the screen drawing/mouse/kb events and helps the speed of a
remote X11 emacs window tremendously.  I recommend it highly.
http://www.xfree86.org/4.2.0/lbxproxy.1.html is the man page for it
(lbxproxy is part of the standard XF86 4.2.0 builds).

-- 
  Ami Fischman
  usenet@fischman.org

Re: Emacs 21/X11 generating unbelieveable network traffic

[Richard Stallman]

    I've got some time on my hands right now, and would like to see this
    problem solved as well, so figured I'd take a crack at it.  d/l'ed the
    emacs cvs repo, looking through it now.  I'm wondering whether there is a
    Roadmap or HACKING-type file for emacs.  Is there some sort of intro doc
    that breaks down the layout of the code?

No, but the code that talks with X for these jobs is mostly in xfns.c
and xterm.c.  Someone already analyzed what causes this slowdown; see
http://mail.gnu.org/pipermail/emacs-devel/2002-September/012633.html.
It is looking up colors and fonts.

With that info it should not take too long for you to find the
relevant code with some breakpoints.  The real job will be designing
a way to avoid doing so many calls to look up colors and fonts.

It would be useful to look further up the call chain and see
what parts of Emacs are doing these lookups.  What is the pattern
of what is going on?

Re: Emacs 21/X11 generating unbelieveable network traffic

[Eli Zaretskii]

> From: Joe Buehler <jbuehler@hekimian.com>
> Date: Wed, 02 Oct 2002 08:33:47 -0400
> 
> I upgraded our internal machines from 20.7 to 21.2 recently.  Most of our
> development machines are on the other end of a 6 mbit link, and boy is
> emacs 21.2 slooooooow.  A couple developers abandoned it for the older version.

There's an entry in PROBLEMS about "Improving performance with slow X
connections"; please see if the advice there helps you.

Re: Emacs 21/X11 generating unbelieveable network traffic

[Kim F. Storm]

Joe Buehler <jbuehler@hekimian.com> writes:

> David Abrahams wrote:
> 
> > I've been trying to figure out why starting Emacs 21 on a remote machine
> > seems so painful. It is  generating reams of network traffic that apparently
> > has nothing to do with graphics. I'm beginning to think that from the
> 
> I upgraded our internal machines from 20.7 to 21.2 recently.  Most of our
> development machines are on the other end of a 6 mbit link, and boy is
> emacs 21.2 slooooooow.  A couple developers abandoned it for the older version.
> 
> It works fine across our local 100 mbit network, so it does look to me like
> 21.2 is a network hog.

Have you tried the advice in PROBLEMS?
In my experience, configure --without-xim makes a significant difference!

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: Emacs 21/X11 generating unbelieveable network traffic

[Richard Stallman]

    It works fine across our local 100 mbit network, so it does look to me like
    21.2 is a network hog.

We are hoping someone who knows a lot about X will offer to work on
this.

Re: Emacs 21/X11 generating unbelieveable network traffic

[Richard Stallman]

    Have you tried the advice in PROBLEMS?
    In my experience, configure --without-xim makes a significant difference!

Is there anyone who can provide code to enable and disable xim at run
time, so that people can obtain this speedup when they want it without
having to forfeit the feature completely or recompile Emacs to use it
again?

Is there a way for Emacs to tell whether the user is using xim
so as to optimize performance when the user is not?

Re: Emacs 21/X11 generating unbelieveable network traffic

[David Abrahams]

"Richard Stallman" <rms@gnu.org> wrote in message
news:E17xad1-0003J7-00@fencepost.gnu.org...
>     Have you tried the advice in PROBLEMS?
>     In my experience, configure --without-xim makes a significant
difference!
>
> Is there anyone who can provide code to enable and disable xim at run
> time, so that people can obtain this speedup when they want it without
> having to forfeit the feature completely or recompile Emacs to use it
> again?
>
> Is there a way for Emacs to tell whether the user is using xim
> so as to optimize performance when the user is not?

The former solution would be an improvement, but probably still very slow
unless there's a command-line option because emacs does a lot of this
network handshaking at startup, which makes it unbearable just to wait for a
prompt.

The latter solution sounds like it might work better.

My intuition, FWIW, is that neither of these is really the right long-term
fix. Even if I was using the "X input method" (I assume that's what xim is),
I don't see why emacs should be shuffling fonts, etc., for every .el file
that gets loaded.

--
-----------------------------------------------------------
           David Abrahams * Boost Consulting
dave@boost-consulting.com * http://www.boost-consulting.com