Subject: Re: x86 instructions reordering
To: Manuel Bouyer <bouyer@antioche.lip6.fr>
From: Jason Thorpe <thorpej@shagadelic.org>
List: port-i386
Date: 03/24/2005 12:46:23
On Mar 24, 2005, at 9:09 AM, Manuel Bouyer wrote:
> And is __insn_barrier() equivalent to x86_lfence() in NetBSD ?
> The generated assembly doesn't change when I use one or the other.
On the i386, __insn_barrier() only performs a memory clobber for the
compiler's benefit (won't re-order around it)... it has no effect on
the CPU's own reordering.
-- thorpej