Path: utzoo!utgpu!watserv1!watmath!att!att!bu.edu!rpi!zaphod.mps.ohio-state.edu!samsung!cs.utexas.edu!evax!utacfd!merch!sneaky!gordon
From: gordon@sneaky.UUCP (Gordon Burditt)
Newsgroups: comp.unix.questions
Subject: Re: Math Coprocessors
Message-ID: <42393@sneaky.UUCP>
Date: 26 Nov 90 22:52:28 GMT
References: <126@mq.com> <1990Nov15.184607.21575@mccc.uucp> <12079@scorn.sco.COM>
Organization: Gordon Burditt
Lines: 20

>    If an application uses floating-point instructions and an FPU
>    is present, then the kernel will make sure the application 
>    uses the FPU.
>    If an application uses floating-point instructions and an FPU
>    is not present, then the kernel will emulate the FPU on behalf
>    of the application.
>    In no case does the kernel itself use floating-point instructions
>    or the FPU.

How, then, is it arranged that process A and process B, both of which use
floating point, don't stomp all over each other's FPU registers?  It would
seem that the kernel must minimally load and unload the FPU state when
necessary due to process switches between processes using the FPU.

If it doesn't use any floating-point instructions, how does it figure out
whether the FPU is present?


					Gordon L. Burditt
					sneaky.lonestar.org!gordon