Re: 4.01 panic/lock issues?

[David Brownlee <abs%NetBSD.org@localhost>]

On Tue, 20 Jan 2009, Izumi Tsutsui wrote:

> abs%NetBSD.org@localhost wrote:
>
> > > Anyway, using -m68040 on building packages is too bad..
> >
> >         I believe that the performance penalty of -m68020 on 040 boxes
> >         is great enough that there is real benefit for using -m68040.
> >         Defaulting to -m68020-60 would reduce that need significantly...
>
> As seen in this thread, -m68040 binaries trigger
> a fatal FPU_EMULATE bug and panics on 020/030.
>
> Even if FPE is fixed, they still cause SIGILL or wrong results
> on 020/030 unless we implement FPE functions for 040/060 specific
> FP instructions. (-m68020-60 is safe of course)

        I think we are in violent agtreement here - we should be
        defaulting to -m68020-60 for userland and kernels which
        need to be able to run on multiple processor types...

--
                David/absolute       -- www.NetBSD.org: No hype required --