Re: Pgcc in Deb
On Tue, Apr 04, 2000 at 12:03:13AM +0300, Eray Ozkural wrote:
> I don't think it's that hard to max out your CPU. The thing is you'll
> notice the difference only when your CPU is maxed out. That's when I
> think [Is this thing compiled with all optimization flags on?]
Excluding Distributed.Net, and my own programs, I don't think my CPU
has been maxed out for more than a few minutes at a time this year. I
keep thinking [how much more would decent SCSI have been?], personally.
> Well, then there's an alternative task that could be done: Making sure
> that all computationally intensive apps compiled with -mcpu=686, since
> anyone using those things will have an 686!!
Nonsense. A lot of people will run a fractal program, or a graphics program
on their home computer which isn't a 686. Some people may even be running
a scientific program on whatever equipment they could scavage, say
an assorted collection of 386's, 486's and 586's.
> > > Are you sure about the PentiumIII? I would think that it's backwards
> > > compatible with PentiumII, and implementing the same optimization paths.
> >
> > The *instructions* are backward-compatible, AFAIK, but the optimal
> > instruction ordering is not. (And ordering for maximum pipelining, more
> > efficient cache usage, optimum memory access, etc., is the major factor
> > in this sort of optimization, not the use of special instructions.)
> >
>
> So the compilers that tuned for 686 won't work as good on PIII, right?
The internals for the PIII are similar enough to the PII that it doesn't make
a difference, IIRC.
> That means the gcc isn't up to date in that, but I always thought the
> gcc
> people worked closely with Intel.
There was a major improvement (funded by Intel) in gcc's Pentium+
optimizations that's in 2.96. (That's the CVS version . . . and it's
not too stable, last I checked.)
--
David Starner - dstarner98@aasaa.ofe.org
Only a nerd would worry about wrong parentheses with
square brackets. But that's what mathematicians are.
-- Dr. Burchard, math professor at OSU
Reply to: