Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 4 Dec 1998 18:34:11 +1100
From:      Bruce Evans <bde@zeta.org.au>
To:        dfr@nlsystems.com, eivind@yes.no
Cc:        freebsd-current@FreeBSD.ORG, luigi@labinfo.iet.unipi.it, tinguely@plains.NoDak.edu
Subject:   Re: 3.0 and double operations in device drivers
Message-ID:  <199812040734.SAA09899@godzilla.zeta.org.au>

next in thread | raw e-mail | index | archive | help
>> > >  > My question is: "Is floating point math now taboo in the kernel?"
>> > >
>> > >  it has always been!
>> > 
>> > I asked for that one :). remove taboo and replace with "forbidden".
>> 
>> It always has been.
>> 
>> (Ie, it has never worked reliably, and this has been a known property
>> of the design).

What design?  Most implementations don't support floating point in the
kernel because supporting would slow down interrupt handling and
context switching.  Since some implementations don't support it, it
should never be used.

>On the alpha, any floating point instructions executed by the kernel
>(except some very careful ones during task switching) should cause a
>panic.

Same for all versions of FreeBSD on i386's, except there are also
some very careful ones in some versions of bcopy() and bzero().
Actually, not so careful.  The panic for unexepected ones went away,
apparently to support unexpected ones in bcopy() or bzero().  See
trap.c rev.1.78.

Bruce

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199812040734.SAA09899>