Date: Wed, 9 Mar 2005 19:59:07 +0100 From: Divacky Roman <xdivac02@stud.fit.vutbr.cz> To: current@freebsd.org Subject: Re: amd64 default CFLAGS Message-ID: <20050309185907.GA10766@stud.fit.vutbr.cz> In-Reply-To: <20050309170100.GG50186@hub.freebsd.org> References: <20050309092749.GA72315@stud.fit.vutbr.cz> <20050309170100.GG50186@hub.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Mar 09, 2005 at 05:01:00PM +0000, Kris Kennaway wrote:
> On Wed, Mar 09, 2005 at 10:27:49AM +0100, Divacky Roman wrote:
> > hi,
> >
> > why is it necessary (if its at all) to have this:
> > -mfpmath=387 -mno-sse -mno-sse2 -mno-mmx -mno-3dnow in default CFLAGS for amd64
> > architecture?
>
> This is the default COPTFLAGS, not CFLAGS, right? You can't use
> special instructions like sse in the kernel because they require extra
> register state operations that would cost performance.
(from sys/conf/kern.mk)
CFLAGS+= -mcmodel=kernel -mno-red-zone \
-mfpmath=387 -mno-sse -mno-sse2 -mno-mmx -mno-3dnow \
-msoft-float -fno-asynchronous-unwind-tables
I'd call it CFLAGS ;)
(from sys/i386/i386/support.s)
ENTRY(sse2_pagezero)
isnt this use of sse in kernel?
why is it allowed in this case and not allowed in general. any measurements how
much does it hurt performance?
roman
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050309185907.GA10766>