Brian Chase <vaxzilla@jarai.org> writes:
> I sort of prefer the idea of pulling the rest of the NetBSD code towards
> something more generic, making all of it less dependent on `n^2' word
> size preconceptions, but I've no sense of how unreasonable that goal
> might be.  Would doing so adversely compromise performance of the kernel
> or, more likely, maintainability of the kernel source?

I prefer to think of it as eliminating assumptions about word length,
endianness, etc. We'll be much more portable to 128 bit systems and
all other sorts of fun as a result. This is likely the only non-n^2
wordlength system we'll ever attempt, so that is not per se the goal
-- the goal is getting rid of assumptions, full stop.

However, we do have to make sure we don't get rid of efficiency in
doing this, yes.

I think the biggest benefit of the effort is a good excuse for
eliminating certain monstrosities. See, for example, the first few
octets of of struct ip.

--
Perry E. Metzger		perry@wasabisystems.com
--
NetBSD: The right OS for your embedded design. http://www.wasabisystems.com/