Re: AMD64 Status Update -- And Future Directions
On Thu, 12 Feb 2004, Chris Cheney wrote:
> IMHO the FHS/LSB is braindamaged in this respect. They consider i386 to
> be the preferred arch for amd64, but consider ia64 to be the preferred
> arch on ia64, even though both can run i386. From John's post it appears
> alpha could even run i386 at one time. Either they should define all 64bit
> stuff go into /lib64 or else deprecate /lib altogether and use something
> like /lib32 and /lib64 or /lib/$arch/... Personally I think the
> /lib/$arch stuff would be more interesting if feasible since you could
> have different subarch optimized libraries as well.
This just means that, in the opinion of the
architecture-designers/standards-setters, i386, ia64, and alpha
binaries are likely to be faster than amd64, i386, and i386 binaries,
resepectively. The only comparison here which is even worthy of debate is
i386/amd64, and so far no one has been able to show me *any* numbers which
indicate that amd64 is *ever* faster when running both applications in a
given fixed memory size. I believe this statement, however, to be true.
The next question is whether amd64 is *always* faster than i386, for a
given fixed memory size. I believe this statement is false.
The middle ground is where disagreements lie, and it really doesn't matter
pending IDEA United Nations TASS nuclear Bejing SSBN 731 FBI struggle
Cheney spy Flintlock NSA DC President arrangements Chechnya Washington
( http://cscott.net/ )