[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
8.0: Timers; Scheduler Sluggishness; Remote Terminal Status Line
Date: Sat, 09 Jun 90 09:21:27 EDT
From: rjb1@gte.com
We've finally upgraded most of our machines to 8.0. Now the
inevitable simple questions begin:
1. Is there any way to get a list of pending timers? The "old
timers" are still in their queue, but the new ones aren't. Even with
the quite lucid documentation of the new scheduler (some Symbolics
tech writer deserves a pat on the back), and with an apropos of the
process package, I can't figure out how to examine all the new timers
to decide whether my nightly utilities are already waiting. I can map
over the processes, but how can I get from a process to its timer
without maintaining my own mapping?
Peek Processes mode has a "Timers" heading; click on it and you'll see
all the new-style timers.
2. Even before I contaminate the system with my kludgey old polled
processes, the new scheduler seems to get horribly bogged down, just
switching among activities. I'm in the habit of simultaneously
firing-up a compile-system or two, starting an individually-filtered
mapped-over move in Zmail, and then getting down to work in Zmacs.
This sort of activity always slowed things down a bit, but now it
seems intolerable. The scheduler seems too "fair" about sharing the
processor among these activities. Has anyone developed a scheduler
policy (is that what I want?) that greatly favors the "current"
activity? (Admittedly, I'm running on a '40; perhaps if we upgraded
our hardware.... But that's "a whole nother" story.)
I've noticed this, too. The old scheduler had per-process quantum
settings; the quantum is the amount of time a process runs before it can
be preempted. Interactive processes frequently got quantum boosts, I
believe, which made them appear snappier.
Unfortunately, there's no documentation on how to extend the "Extensible
Scheduler".
3. Has anyone used the remote terminal on 8.0? Successfully? It
_tries_ to do some really clever things, but none of my terminal
emulators seem to quite cope -- or else it's doing something wrong.
Perhaps I'm overly optimistic to think that a vt320 supports ANSI
X3.<whatever> -- does anybody have a terminal type on which this
works? Or, if that's not the problem, has anyone hacked this code
into submission?
I don't think there have been any major changes to the remote terminal's
use of X3.64 escape sequences in 8.0. It has always had the bug that it
uses "ESC backquote" to move the cursor within the current line, and no
terminal or emulator I know of (except for Symbolics's) supports this
escape sequence (X3.64 is simply a dictionary of control sequences, but
it doesn't specify a required vocabulary, nor is there a way for a host
to find out the supported vocabulary of a terminal).
(For those who haven't tried it, the remote terminal can now include a
status line on the bottom of the screen, not unlike what you see on
the console, complete with an underlined noting-progress -like
indicator. But now, when it gets confused, it sticks in underlining
mode and generously decorates the screen, among other, more familiar
sins.)
I've never seen this. The failure mode I notice is that when command
completion in the CP is rewriting what you've typed so far (e.g. to
capitalize the command name) it does it in the wrong place because of
the above "ESC backquote" problem.
barmar