Re: CPU specific/optimized Debian builds ?
On Thu, May 23, 2002 at 11:29:17PM +1000, Glenn McGrath wrote:
> > > It would be very stupid to suggest it makes it slower.
> >
> > Would it? If you don't have numbers you have *nothing* to back that
> > assertion. We see hand waving advocacy every couple of months about this
> > subject, but in the past several *years* no one has cared enough to do
> > some real research and demonstrate a benefit. "It Stands To Reason" is
> > frankly a ridiculous justification for change.
> >
> > FYI, gcc has a long history of mediocre optimizations, and it
> > is well known that optimizations that make some processors faster can
> > slow down other (even newer) processors (that is, general optimizations,
> > not gcc-specific).
> >
>
> Sounds like your trying to be a troll.
>
> I wonder why CPU manufacturers added extensions to the i386 instruction
> set ?
I present for your general contempt the i486(?) ENTER and LEAVE
instructions, which, at the time of their first appearence, were
slower to execute than the two or three instructions which they were
documented as being exactly equivalent to (and it's far from the only
case of such idiocy from Intel).
Not to mention the nefarious pipeline stall and scheduling issues that
crippled the i586.
--
.''`. ** Debian GNU/Linux ** | Andrew Suffield
: :' : http://www.debian.org/ | Dept. of Computing,
`. `' | Imperial College,
`- -><- | London, UK
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: