Going 64 bit
levi at cold.org
Fri Jan 26 20:41:45 MST 2007
On Jan 26, 2007, at 4:43 PM, Bryan Sant wrote:
> Thanks for the heads up. I assert that Joshua has no sensitivity down
> there and hasn't a clue of what he's talking about. AMD64 is a TRUE
> 64-bit processor, the fact that it can also execute 32-bit
> instructions (natively, there is no emulation), doesn't change a
> thing. Does the fact that your P4 is capable of running a 16-bit DOS
> app natively make it less of a 32-bit proc? The x86_64 instruction
> set, is a true RISC-based, 64-bit, instruction set -- end of story.
Knowing that he's not a complete idiot from conversations with him on
#utah, I decided to ask what he meant. It turns out his objection to
calling it a true 64-bit processor was based on the fact that it
doesn't have a full 64-bit address space. This is not a practical
issue at all, as we aren't likely to be able to make any reasonable
use of a 64-bit address space, so I think his derision of the
processor is more due to his dislike of the x86 family than the
address space issue.
I would also like to see more variety in the computer architectures
of the future, but it doesn't seem to me like the economics will
permit it to happen anytime soon. Witness the attempt Intel made to
switch to the Itanium architecture. It's unusual to see failure of
hardware due to lack of a "sufficiently smart" compiler; that's
usually cited as a reason for software failure!
Anyway, the EPIC idea behind the Itanium is certainly an interesting
one, though I fear that the increasing move towards dynamic languages
could cause even more problems with an architecture that relies so
much on smart compilers.
More information about the PLUG