Asri-unix.555
net.unix-wizards
utcsrgv!utzoo!decvax!ucbvax!hpvax!sri-unix!dave@UCLA-Security
Sun Jan 17 07:44:24 1982
improving terminal response
I'm not sure why you consider the "front end" approach to be "messy".
Each program that wants to be nice just lives in /etc/bin or
somewhere. The front end lives in /bin/<PROGRAMNAME> and does the
nice, and then execs the /etc/<THE COMPONENT LAST OF PATHNAME THE>.
So /bin/edit nices and invokes /etc/bin/edit.
However, I have been considering the following change to my kernel.
I think the problem with editors is that since they do their own
echoing, they must be swapped in to get response. But they are big,
and waiting at a positive (> PZERO) priority, so they are the first
to be swapped out. Fixing the niceness doesn't help here.
My proposal is to fix sched (in slp.c) to modify its computation of
who gets swapped out. The modification would be something along the
lines of:
if the guy is waiting for a terminal, and has been asleep
for less than 5 seconds, he will not be swapped.
else if the guy is waiting for a terminal and has been
asleep for less than 15 seconds, he will be swapped
only if there is nobody else at a positive priority
to get rid of (who doesn't satisfy the same condition).
Comments?
Dave
-----------------------------------------------------------------
gopher://quux.org/ conversion by John Goerzen <
[email protected]>
of
http://communication.ucsd.edu/A-News/
This Usenet Oldnews Archive
article may be copied and distributed freely, provided:
1. There is no money collected for the text(s) of the articles.
2. The following notice remains appended to each copy:
The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996
Bruce Jones, Henry Spencer, David Wiseman.
