Subject: Re: Netscape.
To: Scott Reynolds <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@LAGAVULIN.PDL.CS.CMU.EDU>
Date: 03/12/1995 21:08:54
> > I think that the 4k page size should remain the default for the hp300,
> > as far as what the kernel expects as it's `default' binary format.
> I would tend to agree with Jason here; at the moment, COMPAT_HPUX is in
> need of some attention, and if I understand things correctly the HP-UX
> binaries use 4k page sizes.
There's a big difference between the MMU page size and the "linker page
size". the former is what the MMU sees (and, indeed, for m68k ports is
_only_ 8k for the sun3 and the amiga). The MMU page size cannot be
changed on the hp300 (at least, i don't think it can), because of the
HP MMU used in some of the boxes.
> It seems to be a logical conclusion that
> going to an 8k page would all but kill any chance of HP-UX binaries
Your logic needs work.
It is non-trivial, but possible, to run binaries with a 4k linker page
size on machines with 8k MMU page sizes, and, as noted above, the MMU
page size on the hp300 _shoudln't_ change. It's (as i mentioned) less
efficient in terms of startup time, but i don't consider startup time
a stunning factor the looking at the performance of a daemon.
If it weren't possible to run binaries with a 4k linker page size on
machines with 8k MMU page sizes, i don't think this would have been an
issue at all, and i think the hp300 port would be an m68k8k port now.
the fact that you _can_ do it is the reason (that i've seen) that it's
somewhat acceptable to allow the two different executable formats.