>> It was long time ago I heard developers succeeded in having kerenl
>> profile...
>> 
>        Yes, but the code has been this way since rev 1.1, so if it *ever*
> worked, then there must be something that I'm missing.

It was (one of) my mistake.  Since I started involved with NetBSD/pmax
1.2G, kernel stack usage was altered several times and MCOUNT was not
well maintained to track the changes.

Tohru Nishimura 
Nara Institute of Science and Technology