Re: Version after Version

"Marc Rohloff" <marcrohloff_ng@xxxxxxxxxxx> wrote in message

> Your whole argument is premised on 64 bit operations like
> multiplication. This is a fairly small part of any application. Most
> people don't even use anywhere near the range of 32 bit integers.


Nonsense.
*All* current software can be parallelized to work on 64 bits at the same
time.
You haven't understood basic comp. sci. if you say this.

Simple example: You can do two 32-bit additions in parallel, when doing them
side-by side in a 64-bit register.

And it's extremely naive to focus on things like arithmetic operations on
32-bit numbers.
There is much more than that. There's "operations on data". These operations
happen currently 32-bits at a time, at maximum.

Having the possibility to do double as much in the same number of operations
is what gives the speedup.

Saying that "currently we have not taken advantage of 64-bitness and so we
keep using 32-bits operation"s does NOT prove it's can't be *done* (easily).


> I should also add that larger data size implies more paging to disk.


Ridiculous nonsense.

First of all, who says data size is larger? Why would it be?

*operations* go on twice as much data at the same time, why would the data
itself have to be doubled in size?
You do not understand basic principles of comp. sci. when you say things
like that.

The transition of 32-bits to 64-bits is *huge*.

I wrote in detail why, for database weenies as yourself without basic
understanding of the fundamentals of algorithm design:

http://www.zenhacker.com/TheCaseFor64Bits.htm


.