Sherman.Uitzetter@MAPS.CS.CMU.EDU (06/24/89)
I am using a VAXstation II to write programs for X11R3 using the X toolkit. My question is what is the typical size of such programs? It seems that all (really just two) of the programs I have written come out at around 2 MEGS. Is this normal when the toolkit libraries are used or do I have some unusually large libraries, or what? Thanks in advance for a response. -Sherman.
kit@EXPO.LCS.MIT.EDU (Chris D. Peterson) (06/27/89)
> I am using a VAXstation II to write programs for X11R3 using the X toolkit. > My question is what is the typical size of such programs? It seems that all > (really just two) of the programs I have written come out at around 2 MEGS. > Is this normal when the toolkit libraries are used or do I have some unusually > large libraries, or what? Toolkit applications are big, but not usually that big, here are a few from my home machine (a tek4310): -rwxrwxr-x 1 kit 434176 May 16 13:41 xmh/xmh -rwxrwxr-x 1 jim 418816 Jun 7 16:29 xman/xman -rwxrwxr-x 1 jim 431104 Jun 22 19:31 xterm/xterm -rwxrwxr-x 1 kit 377856 May 11 18:46 xedit/xedit These are all fairly complicated programs, and of course things tend to vary depending upon your machine's instruction set. It sounds like perhaps one or more of your libraries may have been compiled with debugging information. That tends to make toolkit applications really big. Chris D. Peterson MIT X Consortium Net: kit@expo.lcs.mit.edu Phone: (617) 253 - 9608 Address: MIT - Room NE43-213
raveling@venera.isi.edu (Paul Raveling) (06/30/89)
Every layer of support software, including the toolkit, Xlib, and even OS libraries, keeps getting fatter & bringing in what would seem to be more excess baggage. From executable file sizes the tooklit does appear to be the biggest offender, but certainly not the only one. The rest of this message illustrates with some samples. In article <8906271618.AA08184@expo.lcs.mit.edu> kit@EXPO.LCS.MIT.EDU (Chris D. Peterson) writes: > >Toolkit applications are big, but not usually that big, here are a few from >my home machine (a tek4310): Adding sizes for both X11 & X10 on our HP 9000/300 series machines: Tek4310 HP 9000/300 Program X11 X11R3 X10R2 ------- --- ----- ----- >These are all fairly complicated programs, ... xmh 434176 439004 --- xman 418816 410337 --- xterm 431104 423500 188068 ** xedit 377856 385514 --- ** BTW, xterm's level of functional complexity is about the same as what I implemented in less than 32K for a custom terminal a decade ago. That included a small OS kernel. **** Here are some simpler programs: xbiff --- 285780 --- xclock --- 298494 59372 xload --- 289817 46768 **** Here are a couple that don't use the toolkit: xsetroot --- 120308 41052 xrefresh --- 106978 24976 **** And finally, a token example of virtually the simplest possible program, a tiny benchmark, that calls no system functions of any sort, doesn't even include ANY header files in compilation: Assembly language on HP 9000/370: 591 C On an HP 9000/370, compiled under HP-UX 6.5: 5284 C On a NeXt machine, compiled under MACH: 50728 Maybe we should all buy stock in companies that manufacture RAM and disks... or take another look at the architecture of multi-layered system software, including Xlib and Xt. ---------------- Paul Raveling Raveling@isi.edu
jim@EXPO.LCS.MIT.EDU (Jim Fulton) (07/01/89)
> xmh 434176 439004 --- > xman 418816 410337 --- > xterm 431104 423500 188068 ** > xedit 377856 385514 --- > Yes, X11 programs are getting larger, but then they are also beginning to be useful as well (since you're comparing the V10R2 xterm, I assume that it had tek mode in it; I can't remember). It could definitely be smaller, but a fair bit of that overhead is what gives you the ability to tailor the program to your environment as well as make it easier to develop and maintain. For chuckles let's also add a line for the current xterm on a Sun3 with SunOS 4.0.1 and shared Xlib: 327,680 bytes. If Xt, Xaw, and Xmu were shared as well, I'd be even happier (but don't hold your breath, that represents a fair amount of surgery). > ** BTW, xterm's level of functional complexity is about > the same as what I implemented in less than 32K for a custom > terminal a decade ago. That included a small OS kernel. Gee, vt102, tek4014, multiclick selections, 8bit input and output, string insertion, resizing, repainting, dynamic scrolling, colors, popup menus, logging, jump and smooth scroll, all with user-settable bindings in under 32k. > Maybe we should all buy stock in companies that manufacture > RAM and disks... Didn't you know that was what all of these consortia are about! :-) > or take another look at the architecture > of multi-layered system software, including Xlib and Xt. Yes, things have grown (and have even gotten a little fat in places), but functionality and flexibility don't come for free. They're part of what separates hacks from applications (a term which I'm hesitant to apply to any of the programs in the MIT distribution, although xman, xmh, and xterm are starting to come close).