> > I disagree; they should be assumed to work.
>
> I disagree with the law of gravity, myself.

Do you disagree because you find examples where gravity does not work? I
disagree because I have never found an example where dual different
processors don't work, but I've found plenty of examples where they do.

> If it works for you, fine. It certainly isn't something we should ever
> encourage. Lots of stuff breaks -- calibrated loops, applications that
> make use of specific instructions that aren't in both processors,
> etc. If you want to do it, fine, but I don't want anyone wasting time
> making support better for a bad idea.

Why is it a bad idea? Just because you think it's a bad idea?

Code that makes use of calibrated loops should definitely know what
processor it is using. These kinds of loops should only be in the kernel
and drivers, anyway. VAX and SPARC support multiple, different speed
processors, and I've heard of Alphas with different speed processors (but
not running NetBSD). There's no excuse for writing bad code. It's Linux
and FreeBSD coders who write code based upon lame assumptions; if everyone
wrote good, portable code, NetBSD wouldn't be needed as much as it is.

Anyhow, people who compile code with SSE support are very likely to
already know whether their system is a P2 or a P3, or whether it is a dual
processor with both a P2 and a P3.

> > We're NetBSD, and we should be happy to say that we support whatever is
> > thrown our way.
>
> No, we're a volunteer project, and we are happy to include whatever
> support is provided. There is a big difference.

Your distinction isn't very important. One of the nicer selling points of
NetBSD is that we run on just about anything. Unless someone can
illustrate a problem with running two different processors, there's no
good reason to say it's unsupported. Perhaps, for now, one can say it's
not specifically supported, but it certainly works, and one can say that
more testing is a good idea, but it is supported. And while dmesg reports
the two processors as being the same speed when they are not, that is a
bug. If it is a bug that noone cares about, then we should at least be
correct and only report the speed of the boot processor.

John Klos
Sixgirls Computing Labs