🚀 go-pugleaf

In article <kaleb.754677773@kanga.x.org> kaleb@expo.lcs.mit.edu (Kaleb Keithley) writes:
>
>Perhaps someone would like to substantiate heresay with some factual data. 
.
.
>
>I sincerely doubt this had anything to do with Motif. The whole library
>is ~500k on a Sun SPARC. Even if your link editor pulls in every last
>module in libXm, libXt, and libX11, (statically linked) the libraries
>only contribute ~1.5meg, unstripped. That 35 meg came from somewhere
>else.
>
>Kaleb Keithley
>

     Certainly the library is not all that large, but the library size has
little to do with the actual amount of memory used at run time.  

     I do not have the exact numbers available, but I recall that the Motif
list widget (in 1.1.x) used an exorbitant amount of memory.  I did a project
that used very long scrolled lists (30 k lines).  These lists would take up
6 to 8 meg of memory in the list widget itself.  (I haven't tried this with 1.2.x
so maybe its fixed now.)

     So, the impression of Motif being bloated is not completely off base as 
far as I am concerned.  It wouldn't take alot of long list widgets that the 
gui builder creates to add up to 35 meg.  The program that I worked on which
included these long lists routinely ran at 20 megs.

     Just my 2 cents worth.

Glenn Pinkerton
glp@csn.org

In article <kaleb.754677773@kanga.x.org>,
Kaleb Keithley <kaleb@expo.lcs.mit.edu> wrote:
>I sincerely doubt this had anything to do with Motif. The whole library
>is ~500k on a Sun SPARC. Even if your link editor pulls in every last
>module in libXm, libXt, and libX11, (statically linked) the libraries
>only contribute ~1.5meg, unstripped. That 35 meg came from somewhere
>else.
>
	Hmm, was this runtime or load time sizes? Up until the current release
	Motif had/has a tendonsy to leaked memory like a seive. B^(. This was
	especially painful on small PC systems running SCO or Interactive
	although it would eventually be a pain on Sun's, HP's and RS6k's.

		-Rob
#include <std/disclaimers.h>
Speaking for myself only of course.

In article <kaleb.754677773@kanga.x.org>, kaleb@expo.lcs.mit.edu (Kaleb Keithley) writes:
|> Perhaps someone would like to substantiate heresay with some factual data. 
|> At my prior employer I used Motif to develop a "real-time" status display 
|> to monitor ground equipment. The programs responded instantaneously and
|> repeated profile analysis showed that the negligible amount of CPU that 
|> they did use was not in Motif. I would characterize Motif as anything but 
|> slow.

Yah, with a workstation on my desk, its fine.  However, its widgets do tend
to redraw themselves entirely, say, rather than just the part that gets
uncovered.  Nothing quite like running a Motif app over a SLIP connection
with a line speed of 9600 (or even 38400, for that matter).

But that's neither here nor there w/regard to using it for free software.

-- 
: jay          jay@vnet.ibm.com    My opinions and ideas, not my employer's.
: shmdgljd@rchland.vnet.ibm.com    (c) Copyright 1993.  All rights reserved.

In article <kaleb.754677773@kanga.x.org>, kaleb@expo.lcs.mit.edu (Kaleb Keithley) writes:
|> Why is this cross-posted to so many groups?
|> 
|> deal@ait.nrl.navy.mil (Butch Deal) writes:
|> >In article Frank.Peters@MsState.Edu (Frank Peters) writes:
|> >|> In article <CH9Jnu.M54@uceng.uc.edu>
|> >|> 	gjunker@ddt.eng.uc.edu (Gregory Junker) says:
|> >|> 
|> >|> >	- Slow
|> >|> 
|> >|> I doubt it.  You may have a slow X server or slow hardware or a slow
|> >|> Motif application (or a slow Motif window manager, which is just another
|> >|> application).  I've seen and heard nothing to suggest that the Motif
|> >|> toolkit inherently makes your applications slow.
|> >I agree.  It is slow in the sence that it wastes alot of cpu time.  Just
|> >because you have a fast computer, you do not want to waste all the cpu time
|> >with a GUI.  Motif is overly bloated
|> 
|> Perhaps someone would like to substantiate heresay with some factual data. 
|> At my prior employer I used Motif to develop a "real-time" status display 
|> to monitor ground equipment. The programs responded instantaneously and
|> repeated profile analysis showed that the negligible amount of CPU that 
|> they did use was not in Motif. I would characterize Motif as anything but 
|> slow.

a comparison of two programs like xnet vers. 1.3 and mosaic.  Actually doing a
trace with xnet and just moving the pointer around in the mosaic window. 
ps -auCS on a sun690 MP (super sparc) with X11r5 (patch 26) and piewm with 
raisedelay patch installed.

USER       PID %CPU %MEM   SZ  RSS TT STAT START  TIME COMMAND
deal      7890  0.0  3.8 3600 4668 co S    Nov 29 14:41 /usr/bin/X11/Xsun :0 -au
deal     16121  0.0  0.4   80  540 p2 S    14:09   0:00 /usr/local/bin/xnet
deal     16123  0.0  0.0 1308    0 p2 IW   14:09   0:02 Mosaic

xnet trace to sun.com (9 hops away from here)
just moveing the pointer in the mosaic window.

USER       PID %CPU %MEM   SZ  RSS TT STAT START  TIME COMMAND
deal      7890  0.8  3.8 3600 4668 co S    Nov 29 14:45 /usr/bin/X11/Xsun :0 -au
deal     16121  0.0  0.4   80  544 p2 S    14:09   0:00 /usr/local/bin/xnet
deal     16123  0.0  0.7 1308  792 p2 S    14:09   0:03 Mosaic

bring up help and info windows of xnet and another trace to sun.com
move pointer around some more in mosaic window.

USER       PID %CPU %MEM   SZ  RSS TT STAT START  TIME COMMAND
deal      7890  0.0  3.8 3600 4668 co S    Nov 29 14:50 /usr/bin/X11/Xsun :0 -au
deal     16121  0.0  0.4   84  544 p2 S    14:09   0:00 /usr/local/bin/xnet
deal     16123  0.0  0.8 1308  936 p2 S    14:09   0:04 Mosaic

three traces with xnet and a source route trace, help 5 times
get a local home page for mosaic

USER       PID %CPU %MEM   SZ  RSS TT STAT START  TIME COMMAND
deal      7890  0.0  3.8 3600 4668 co S    Nov 29 15:01 /usr/bin/X11/Xsun :0 -au
deal     16121  0.0  0.5   88  552 p2 S    14:09   0:00 /usr/local/bin/xnet
deal     16123  0.0  0.8 1416  948 p2 S    14:09   0:05 Mosaic

This is also with a local console.  The network load becomes very apparent
with motif if you are useing a remote computer and have a heavy  ether load,
or a PPP link.

|> 
|> >|> >	- Difficult (read impossible) to program
|> >|> 
|> >|> I can't really argue this since I've never tried to write a Motif program.
|> >|> But the various applications out there seem to argue against this.
|> 
|> You obviously haven't tried to program in the raw Xlib, Macintosh,
|> Presentation Manager, or Windows environments.

I didn't write that motif was difficult to write in, but I have writen Xlib, 
XT, Mac, Motif, OpenLook, Xview, and suntools programs.  Motif is not
particularly harder than any of the rest although I think the docs are a bit
harsh.  I do much prefer Xlib though over the others, although Xview is very
easy to come up with a quick easy application that looks nice.

|> 
|> >Many of them were writen with a GUI tool.  These GUI tools, although very easy
|> >to use, yeald very hefty programs.  I have seen a 35Meg motif binary (striped).
|> >This is not always the case in any means.
|> 
|> I sincerely doubt this had anything to do with Motif. The whole library
|> is ~500k on a Sun SPARC. Even if your link editor pulls in every last
|> module in libXm, libXt, and libX11, (statically linked) the libraries
|> only contribute ~1.5meg, unstripped. That 35 meg came from somewhere
|> else.
|> 
I will not debate this point much as I did not look at the source code for
this program.  I think almost everyone has seen how much larger a motif
program is than an Xlib or Xt program that does the same.

|> --
|> 
|> Kaleb Keithley
|> 

-- 
#include <std/*>
The Butcher                  
Butch Deal                   deal@ait.nrl.navy.mil
--------------------------------------------------------------------------------

glp@teal.csn.org (Glenn Pinkerton) writes:

>I wrote:
>>I sincerely doubt this had anything to do with Motif. The whole library
>>is ~500k on a Sun SPARC. Even if your link editor pulls in every last
>>module in libXm, libXt, and libX11, (statically linked) the libraries
>>only contribute ~1.5meg, unstripped. That 35 meg came from somewhere
>>else.
>>

>     Certainly the library is not all that large, but the library size has
>little to do with the actual amount of memory used at run time.  

Who said anything about runtime? The original post referred to a 35 meg
stripped executable *FILE*.

Think and read before you post. Sheesh.

In article <kaleb.754689986@kanga.x.org>, kaleb@expo.lcs.mit.edu (Kaleb Keithley) writes:
|> glp@teal.csn.org (Glenn Pinkerton) writes:
|> 
|> >I wrote:
|> >>I sincerely doubt this had anything to do with Motif. The whole library
|> >>is ~500k on a Sun SPARC. Even if your link editor pulls in every last
|> >>module in libXm, libXt, and libX11, (statically linked) the libraries
|> >>only contribute ~1.5meg, unstripped. That 35 meg came from somewhere
|> >>else.
|> >>
|> 
|> >     Certainly the library is not all that large, but the library size has
|> >little to do with the actual amount of memory used at run time.  
|> 
|> Who said anything about runtime? The original post referred to a 35 meg
|> stripped executable *FILE*.
|> 
|> Think and read before you post. Sheesh.
|> 
yes the file was 35 Meg but the machine it was run on (a sparc 2 64 Meg Ram)
needed 200 Meg swap to run it.  There were alot of menues and lists.
The program did have a large pixmap compiled in but not THAT large.
-- 
#include <std/*>
The Butcher                  
Butch Deal                   deal@ait.nrl.navy.mil
--------------------------------------------------------------------------------

In article <kaleb.754677773@kanga.x.org> kaleb@expo.lcs.mit.edu (Kaleb Keithley) writes:

>   Perhaps someone would like to substantiate heresay with some factual data. 
>   At my prior employer I used Motif to develop a "real-time" status display 
>   to monitor ground equipment. The programs responded instantaneously and
>   repeated profile analysis showed that the negligible amount of CPU that 
>   they did use was not in Motif. I would characterize Motif as anything but 
>   slow.

>   Kaleb Keithley


  Just wondering whether you have tried XmText -widget with word wrap on.
  Well, you just need to wait a few seconds when you type something to see
  what you typed. And this is on SPARCstation.
  


						Karri