[Info-vax] OpenVMS I64 V8.1 "Evaluation Release"?

[John Wallace]

On Mar 19, 9:23 pm, glen herrmannsfeldt <g... at ugcs.caltech.edu> wrote:
> Dennis Grevenstein <dennis.grevenst... at gmail.com> wrote:
>
> (snip, someone wrote)
>
> >>> I sometimes wonder what might have happened if AMD
> >>> had not come up with x86_64.
> >> That's like speculating how history would have developped
> >> in a parallel universe.
> > Sure, but IA64 had a chance to take over serious x86 market share,
> > because the old x86 was only a 32bit platform.
>
> Yes, but how many really need 64 bits? How many computers sold
> today have more than 4GB of RAM?
>
> Much of high-end scientific computing has been made affordable
> by people buying processors that they didn't need, driving the
> economy of scale.
>
> Even more, IA32 with a segment descirptor cache would have allowed
> addressing over 4GB pretty easily. While some programs do need more
> than 4GB, not so many need so much in one contiguous addressable block.
> The IA32 MMU doesn't allow for more than 4GB (mistake one), and,
> without a descriptor cache the overhead of loading segment descriptors
> would slow down large model 32 bit code way too much.
>
> The small number of registers in IA32 was also a problem, and
> x86_64 does help there. That might have been fixed without going
> to 64 bits, though.
>
> > I'm not saying that VMS would be much more popular. I'm just
> > saying that it wasn't nessecarily a bad idea to get into the IA64
> > market back then.
>
> For the real high-end servers that really need the large address
> space, IA64 might not have been so bad, but it seems to me that it
> took too long to get to market, partly because of the need for IA32
> compatibility.
>
> -- glen

Various folk seem to be confusing various related concepts here, not
specifically Glen.

32+ bit physical memory does not require 32+ bit logical address -
various Pentium/Xeon with the PAE feature could have more than 32bits
worth of physical memory, they just couldn't access it all at once.
(Some readers will remember that the PDP11, with a 16bit virtual
address, could similarly quite happily address 22bits worth of
physical memory on some models)

64bit application design does not require >4GB of RAM, though more RAM
may often be helpful for performance. But even with less than 4GB of
memory, it's often easier for some apps to be designed to assume a
64bit virtual address and let a decent OS worry about mapping virtual
memory to physical memory.

"If one really desperately needed >4GB, one could just as well have
chosen one of the mature 64bit RISC platforms."

Such as? With what OS? If you really wanted flat 64bit addressing, for
a long time Tru64 was the answer (even if it wasn't called that at the
time).

"If accessing more than 4G of RAM was an issue, surely that can be
done without 64 bits."

Exactly. But doing it *simply* wasn't always easy, and doing things
the "easy" way ("the software crisis") was starting to become more
important than doing things well or cost effectively.


"I'm just saying that it wasn't nessecarily a bad idea to get into the
IA64
market back then. "

There never was "an IA64 market". It looks like there was a private
agreement between toplevel board members in two companies to stop
competing with each other in certain important areas.

"Itanium is only alive because Intel has alot of money to burn. "

That, plus the fact that Intel *really* needed to show they know how
to do things outside the x86 market. They still do.