29 Aug
2006
29 Aug
'06
3:56 p.m.
der Mouse wrote:
Yes, but this is an artificial distinction. Main memory is
still homogenous to the CPU/programmer.
Some ideas
probably *should* have been lost (e.g., Harvard
architecture, non-binary radixes, etc.)
Harvard architectures are not dead. If you look at just the CPU core
and cache of a modern machine, you have a basically Harvard setup; it
doesn't become von Neumann until you lump the CPU and cache together as the processor and go to main memory. Harvard
architectures are also
used in specialized applications, such as DSPs and microcontrollers.
Non-binary radixes...I'm not sure I think those are better lost. Base
[snip]
Sorry, I wasn;t clear in my original comment. What I meant was
"non binary -- 8421 -- encodings". E.g., gray code (though useful
for shaft and other incremental encoders), bi-quinary, 5421,
2-of-5, etc.
While some like Chen-Ho and DPD are interesting hacks, I still
haven't found them to be particularly *useful* -- from a
software standpoint. (though they can have use in specialty
applications -- often with specialty hardware and/or
protocols).
I'll "hold out" on BCD and packed BCD; and will concede
that there are times when sign-magnitude can be a win.
But, most of the other encodings are just... *icky*! :>