Date: Mon, 9 Feb 1998 09:54:52 +1100 (EST) From: Andrew Reilly <reilly@zeta.org.au> To: tlambert@primenet.com Cc: dap@damon.com, nate@mt.sri.com, hackers@FreeBSD.ORG Subject: Re: MMX, pentium, etc Message-ID: <199802082254.JAA03377@gurney.reilly.home> In-Reply-To: <199802080736.AAA28517@usr02.primenet.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 8 Feb, Terry Lambert wrote: > MMX resuses the FP registers. This is because they couldn't define new > registers because new registers would not get saved and restored on > process context switch for any existing OS's. > > Thus if you use FP, you can not use MMX, and vice versa. MMX and the virtues of integer math aside, I have a question for the kernel hackers: Just how hard would it be to be able to build a special case FreeBSD kernel that knew how to save some extra registers at a context switch? Aparently there's to be a new version of the IDT Centaur C6 later this year that allows you to map the floating point stack into seven of about _30_ real, indexable floating point registers, and matches this with pipelined, single-cylce-dispatch multiply-accumulate instructions. For the sort of work that I do, this would be the absolute bees knees. Acording to Byte, they plan to get around the problem of Windows (NT) not saving the extra state by making sure that you only use it inside their Direct3D libraries, but that seems like a dumb option for a system like FreeBSD that is fully re-compilable, and therefore readily able to take advantage of developments like this... -- Andrew "The steady state of disks is full." -- Ken Thompson To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199802082254.JAA03377>