Re: Report on NetBSD-CURRENT and `hpcmips` on MobilePro 780

On Fri, 22 Dec 2023 at 20:55, Warner Losh <[email protected]> wrote:

> On Fri, Dec 22, 2023, 3:55 PM Alexander Bochmann <[email protected]> wrote:
>
> > ...on 2023-12-21 11:47:28, David Brownlee wrote:
> >
> >  > It should be possible to build a soft-float userland which avoids FPU
> >  > entirely. If the relevant issues go away then that isolates it to an
> FPU
> >  > issue
> >
> > Oh well... The detail I forgot about since last looking at this
> > is that the build configuration in sys/arch/hpcmips/conf/std.hpcmips
> > has
> >
> >  > options         NOFPU                   # No FPU
> >  > options         FPEMUL                  # emulate FPU insn
> >
> > ...since there's no FPU in the vr41xx CPUs.
> > (Insert facepalm emoji here.)
> >
> > From a quick grep at least, no other MIPS sub-architecture seems
> > to use that option, so probably no one else would notice if
> > something was broken in that code. Then again, that could be
> > another ill-advised train of thought.
> >
>
> You could do a nice a/b test. Boot a r4000 Malta in qemu.
>  Confirm it works.
> Add the options to the Malta kernel and try again.
>
> Warner
>
> I just did the a/b test. The MIPS CI20 works with or without these
options.
This is on head + the patch from
https://mail-index.netbsd.org/port-mips/2023/07/07/msg001290.html