Re: Speaking of PAE

[John Nemeth <jnemeth%cue.bc.ca@localhost>]

On Jan 29, 10:23pm, Martin Husemann wrote:
} On Fri, Jan 29, 2016 at 02:09:17PM -0700, Swift Griggs wrote:
} > 
} > Can any of the kernel gods comment on if PAE is still considered unstable? 
} > It's hard to tell if it's disabled by default because it conflicts with 
} > the Xen memory model, or if it's because it causes stability issues.

     PAE is required for Xen, at least for dom0.  BTW, Xen questions
are best addresses to port-xen.

} Not unstable (AFAICT), but it comes with a performance penalty.
} 
} In some environments it makes sense (e.g. a i386 xen domU on a 64bit
} setup and little virtual memory allocated - I'm running one of those
} myself), but in most cases a straight i386 kernel (for max 4GB
} installed) or a straight amd64 kernel otherwise are preferable.
} 
} But if you go with a amd64 kernel, you still can install i386 userland
} and have all the same benefits you get with i386-pae. Modulo bugs in the
} compat_netbsd32 emulation the userland side of things should be the same.

     Also, differences in device numbers.  Be sure to use an amd64
version of MAKEDEV.

} There are a few minor issues (e.g. i386 savecore can not deal with a amd64
} kernel crash dump). We use this type of setup as default on mips64 and
} are trying to minimize the fallout.
} 
}-- End of excerpt from Martin Husemann