On Jan 1, 8:58 pm, Carnations <Beauti...@Carnations.com> wrote:
> I fail to see why 64bit applications should perform "better" than their 32bit counterparts...
(We often get this 64 vs. 32 bit argument in nz.comp, and I've not
seen this point raised here, so I'm leaping in with it somewhat
arbitrarily here.)
Using a large address space, much larger than the hardware addressable
memory, can be used for different programming paradigms. An early
such approach was mapping files (f.ex. mmap) instead of the open,
read, write stuff, and another doing communication that way. Later,
one maps resources on the internet into one's address space. Let the
OS worry about protocols and APIs and security, just do stuff with
memory, or some abstraction like objects. Or, imagine a functional,
lazy evaluation view of the universe.
It's not necessarily a better way, and it can break the Unix
everything-is-a-file concept. But, for some applications, it can be
vastly simplifying, and for some IO bound tasks getting the virtual
memory system to do all the IO makes them much faster.
I'm sure there's much more eloquent and useful descriptions of these
ideas around, this is off the top of my head. I suppose only big
servers use these approaches presently, and I suppose not many of them
use 32 bit addressing.
Regards, John
|