Ultrix 3.1 (aka UWS 2.1) memory requirements

[Ken A. Nishimura]

In article <GRUNWALD.89Aug2153953 at flute.cs.uiuc.edu> grunwald at flute.cs.uiuc.edu writes:
>
>How much does DECwindows take compare to e.g., the MIT X11R3 distribution?
>Why does it take more?
>
>Since the monochrome PMAX doesn't have a hardware accelerator, I assume
>that there's little performance improvement in using DECwindows over
>X11R3; am I right?

Well, the problem with DECwindows is worse on systems with little
core memory.  If there is one word to describe DECwindows or
as a matter of fact any /usr/bin/dx*, it is HUGE.  These are huge
binaries, due to the enormous libraries associated with DECWindows.

When running DECwindows on a memory starved machine, raising a window
or doing any kind of context switch is painful, since it swaps anytime
you do something.  Plus, it costs a lot in terms of swap space.  We
found that we had to add another 12MB of swap to accomodate worst
case DEC* usage.

MIT X11R3 binaries are smaller, so the chance of things getting
swapped out at every touch of the keyboard is smaller, plus it
saves on virtual memory....plus I don't think it is as CPU intensive
in the stuff it does behind the scenes, though I'm not sure about
this.

As for the PMAX, the color PMAX has no graphics accelerator either
(e.g. no dragon chips).  The graphics performance of the PMAX is
entirely dependent on the R2000a processor.  When unloaded, the PMAX
will edge out the GPX, but when loaded, the display suffers.

The problem with DECW* is worse on PMAXen for two reasons.  (1)
The binaries are larger.  (2)  DEC didn't use shared libraries....
Hence, every single /usr/bin/dx* application is 1MB of text.  It
gets really expensive after a while, especially if you are on
an 8MB PMAXen.

				-ken