Date: Wed, 25 Nov 1998 15:01:13 -0500 (EST) From: Alfred Perlstein <bright@hotjobs.com> To: Marius Bendiksen <Marius.Bendiksen@scancall.no> Cc: ben@rosengart.com, hackers@FreeBSD.ORG Subject: Re: FreeBSD on i386 memory model Message-ID: <Pine.BSF.4.05.9811251454550.10312-100000@bright.fx.genx.net> In-Reply-To: <3.0.5.32.19981125194556.00998cb0@mail.scancall.no>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 25 Nov 1998, Marius Bendiksen wrote: > >Why don't you implement it and find out? > > (1) I lack the required knowledge of BSD internals > (2) I doubt I have sufficient mastery of C to rewrite libc > (3) If the effort is going to be wasted (ie. if noone will commit it), > there's no point in doing so. > > Correct at least two of the above, and I'll give it my best shot, and get > back to you. I assume it would be -quite- a lot of work? > The syscall entry into the kernel is done via a macro afaik, redefine it, and code the alternate entry into the kernel. If you wanted a libc that you could "move" between machines..., you could recode the libc bootstrap to sysctl the processor model number and provide a jump point to either entry point. However, you then slow down both entry methods by the cost of an additional jump/return and you cause the instruction cache to be less effective as you jump execution around in memory. Another idea is 2 lib directories, one optimized for <= 486 and the other optimized for > 486, then you double the amount of space you need for shared libs... Not worth the cycles unless you were doing long term modeling, and then, as asked before, why use a 486? It could be made into a Makeworld option perhaps? -Alfred To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9811251454550.10312-100000>