While cleaning up my mailbox, I found the following:

Excerpts from netbsd: 1-May-97 Re: 2.0beta2 bootblocks on .. "Perry E.
Metzger"@pierm (517)

> Jonathan Stone writes:
> [...]
> > One thing I've noticed is that the 2.0beta bootblocks seem much slower
> > at reading a kernel off a floppy than the old 1.27 (or thereabouts)
> > bootblocks. Has anyone else seen this, or is it just me?
> 
> Its possible. We did add some performance hacks to try to read a track
> at a time (at one point it was *VERY* slow), but...


Is this really still slower than the old boot - can you measure it?
I've tested all the libsa UFS and the i386 biosdisk functions in a
user space environment some weeks ago and it showed that
the read-ahead works properly - almost everytime whole tracks
are read, and they are read only once.
I can't imagine that processing overhead causes a slowdown
on a reasonable machine.

best regards
Matthias