Re: Version after Version


"Frank de Groot" <franciad@xxxxxxxxx> wrote in message
news:gQ39f.389$3O1.15023@xxxxxxxxxxxxxxxxx

> It is sad that most software developers, be it old or young, do not
> understand the incredible speedups possible with 64-bit processing. Such
> speedups often far exceed a factor of two. Obviously, no amount of
> explanation will make you understand. It might have something to do with
> a lack of formal education on how a computer works, if you are 49, you
> migh not have had the advantage of a comp. science foundation. Still, I
> think it should be possible to grasp the essentials of the great
> possibilities 64-bit processing offers to speedup code. This is exactly
> why MMX was invented, but "pure" 64-bit processing ascends MMX, as it
> offers conditional branching, amongst other things.

64-bit computers are nothing new. DEC had the, IIRC, DEC-10 and there was
also the . . . da** the name escapes me at the moment. In any case while
the machines had some commercial sucess, mostly in university and
scientific environments, they did not do well in business environments. We
do, as a community, have quite a bit of experience working with 64-bit
machines. It is known fact that bigger word size can improve performance of
some types of computation, just as it can slow down other types. Its also
known fact that bigger word size does tend to produce programs with larger
memory requirements.

The devil is really in the details. That is, how the cpu actually
implements its instruction set and the throughput of the cpu-memory bus.

In my experience there is usually a penalty to pay for part-word work. So,
while a program may only need to work with 32-bit integers, it may well
prove more efficient to use 64-bit ones. Hence the argurment that 64-bit
architectures use more space.

I would also point out that while modern desktops tend to process much more
non-textual data than those of even a few years ago, they still process
vast amounts of text. No matter how you cut it, 64-bit word size does not
improve text handling. In fact, it may well haper it, depending of course,
on implementation.

Until we actually have wider experience with 64-bit Windows platforms being
used as desktops its going to be tremendously difficult to judge how
beneficial the architecture may be. Personally I doubt that its going to
make a real difference to the "average" desktop. Simply because cpu -
memory - disk throughput will continue to be the real limit on system
performance.


.