Re: program launch time improved?

[David Holland <dholland-current%netbsd.org@localhost>]

On Sun, Jan 22, 2012 at 08:09:21PM +0000, bch%methodlogic.net@localhost wrote:
 > > > Has a disk, io, or some other subsystem been affected lately that would
 > > > drastically improve program loadtimes from a freshly booted system (I'm
 > > > amd64). Over the last few weeks, when launching my typical apps after
 > > > a fresh system boot (Firefox, emacs, tclsh...) the load times until
 > > > a display/s were l-o-n-g. Typical scenario would be what appeared to
 > > > be an inordinate amount of disk activity before program launch. Since
 > > > ~yesterday, it seems things are *dramatically* faster.
 > > > 
 > > > Apologies for the really subjective mail, but am curious what's changed,
 > > > and this description may be enough for the people who changed [whatever]
 > > > to say "That sounds like something I could have affected." Whoever the
 > > > responsible people are, keep up the excellent work!
 > > 
 > > The only thing I recall seeing was silly locking in fileassoc... so
 > > maybe if you have that turned on, that's it?
 > 
 > It's in my kernel config (which is custom, but indeed GENERIC has it
 > too), so perhaps? Were those changes (to bad locking) yours David? Would
 > you expect to see a dramatic improvement? To my mind, the perceived
 > improvement is just short of mind-blowing. To be fair and honest though,
 > the prev. performance was horrible, in my opinion.

It was unnecessarily taking kernel_lock; that doesn't really match the
symptoms you describe, but again, it was the only thing I can recall
seeing. (no, it wasn't my commit)

-- 
David A. Holland
dholland%netbsd.org@localhost