[5031] in Athena Bugs
Re: maxtextcache
daemon@ATHENA.MIT.EDU (John F Carr)
Fri Jun 1 20:10:47 1990
Date: Fri, 1 Jun 90 20:10:26 -0400 (EDT)
From: John F Carr <jfc@ATHENA.MIT.EDU>
To: probe@MIT.EDU
Cc: release7@MIT.EDU, testers@MIT.EDU
In-Reply-To: <9006012347.AA22245@odie.MIT.EDU>
See rel-eng meeting transactions 2354, 2353. In these Ezra says the
reason was related to performance.
That the cache routines were flawed does not mean that the text cache
should be disabled (I have talked to Ed and am aware of the nature of
the problem). It means that maxtextcache should be set either to 0 OR
-1. At this stage in the release, it is my understanding that the
usual solution to a problem without a quick fix is to revert to the
previous release behavior. The last release had maxtextcache == -1.
Since we have the rmalloc change to free unused texts when swap space is
exhausted, much of what the maxtextcache reduction was intended to fix
has already been fixed (especially considering that enlarged swap on RTs
is considered a prerequisite for this release). Setting maxtextcache to
0 certainly will hurt performance (worse than it is with maxtextcache ==
10 in the current kernel).
There is a mailing list where kernel problems can be discussed. This
would have been a good use for it.
