Re: Windows XP 64

Arny Krueger <arnyk@xxxxxxxxxx> wrote:
"Laurence Payne" <NOSPAMlpayne1ATdsl.pipex.com> wrote in
message news:68lvl39h60ouvmpktkrl3i7pmvl7ut0ce3@xxxxxxx
On Wed, 12 Dec 2007 07:30:05 -0500, "Arny Krueger"
<arnyk@xxxxxxxxxx> wrote:

The "bit size" of a computer refers to largest datum it
can handle in a single instruction. It has nothing, per
se, to do with the size of the address space.

Actually, that hasn't been true for about 40 years.

Back in the late 60s, just about any self-respecting
computer would support single precision floating point,
which was around 32 bits

But could it manipulate it in a single instruction?

Yes.

There were even mainstream business computers like the CDC Cybers, that had
*only* 60 and 120 bit data words. Nothing shorter.

The Cyber was an abomination.
It had only 60-bit data words, true. But, they were only floats. If you
wanted to use an integer, you treated it like a float with a zero exponent.
So a single precision float was 60 bits, but an int was 48 bits.

Addresses were only 18 bits long, and if you moved an integer to an
address register it used only the lower 18 bits. Oh yeah, and since there
was no virtual memory, you're stuck with those 18 bits.

They did some trickery on the Cyber 180 series to allow virtual memory,
but nobody used it because it took forever for CDC to get their virtual
memory OS shipped, and when they finally did it was a bloated monstrosity
that nobody wanted. Sort of like TSS/370.

It's the same issue with addressing. Sure, a 32-bit system
could address any memory size you liked. In a sense, it
does it whenever it reads a large hard drive. But could
it do it quickly, without a paging or indexing layer
getting in the way?

Yes. The IBM 360 instruction set had a full complement of instructions for
32, 64, 128 bit floating point, and up to 2k bit character data. The 360s
(other than the 67) were real memory computers, so no paging. The
instructions supported indexed addresses, but the address index registers
were static during instruction execution so there was no game-playing there.

Well, you could throw a DAT box on your 360/50... As I recall the 360
also had 16-bit addresses but it's been 30 years since I wrote an RS
or an RX...
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
.

Relevant Pages

  • Re: [Lit.] Buffer overruns
    ... > floating point support or a memory expansion option. ... had virtual memory support grafted on. ... > where the modified instruction was fetched from. ... vis-a-vis the official coporate strategic operating system TSS/360. ...
    (sci.crypt)
  • Re: [PATCH] Mantaining turnstile aligned to 128 bytes in i386 CPUs
    ... :This doesn't contradict your claim since main memory is not really involved. ... that gives the same not-very-real-world cache state for all iterations ... full, and the cpu stalls anyway. ... static instruction order makes it easiest for them, ...
    (freebsd-arch)
  • Re: [PATCH] Mantaining turnstile aligned to 128 bytes in i386 CPUs
    ... :This doesn't contradict your claim since main memory is not really involved. ... that gives the same not-very-real-world cache state for all iterations ... full, and the cpu stalls anyway. ... static instruction order makes it easiest for them, ...
    (freebsd-current)
  • Re: C and Low-level Storage
    ... stored in addressable memory. ... can only be read for execution. ... CPU has a an instruction fetch unit that reads the instruction data ... The closest examples I can thing of are some analog computers. ...
    (comp.lang.cobol)
  • Re: New ARM Cortex Microcontroller Product Family from STMicroelectronics
    ... FIFOs or instruction prefetching are not ... It is much better to just have single cycle ... memory. ... The Stellaris CM3 running at 50 ...
    (comp.arch.embedded)
Loading