Subject: Re: m68k soft float (was illegal instruction)
To: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
From: John Klos <john@sixgirls.org>
List: port-mac68k
Date: 03/05/2003 11:56:14
Hi,

> All m68k ports share most binaries on the release. I don't like to use
> soft-float binaries on news68k because all news68k machines have FPU. No
> softfloat binaries are needed at all since they makes slow machines
> slower.
>
> If you really want soft-float binaries, we should provide both
> hard-float and soft-float binaries and make soft-float ones optional.
> For shared libraries we can use ld.so.conf(5) to switch shlibs
> by sysctl value, but I don't have any idea for static binaries.

I don't know if I gave the impression somehow that I thought we'd just
switch everything to soft float, but that is not not case. When the code
is committed, it will allow for people who put MKSOFTFLOAT=yes in their
/etc/mk.conf file to build soft float stuff. Obviously, you don't have to.

> Anyway, implementing proper FPE is the real solution, I think.

The issue here is that the FPE is useless because of an uncorrectable bug
in the chip.  There's no way around that bug without significant overhead.

> > First, let's get the code committed and well tested. Then we can talk
> > about how to deal with the rest. Yes?
>
> No. You should ask all m68k port-maintainers first.

I will certainly give them a heads up.

John Klos
Sixgirls Computing Labs