> So either you do not understand my position, or you are
> disingenuously twisting my consumer oriented position to one
> of an engineering point of view. Or maybe you just don't
> understand the difference between someone buying a laptop for
> home use and someone needing a laptop to write and compile next
> generation applications (that is your job, right?).
I'm trying to get you to understand that VM is really not that
bad. But address space can be really, really useful. That's
pretty easy for someone with a CS degree to understand.
Optimizing software is only a hobby for me.
> Yes, yes... but as I already pointed out, the same can be
> said for dual core technology. Buy a single core Turion
> today, and you miss out on all the performance prospects
> of multithreaded environments, which benchmarks have shown
> can already offer dramatic improvements.
In some applications. Will the typical user see those improvements in the workloads that they run, though?
BTW, I'm of the mind to wait for the DC Turion or buy a
DC Athlon X2 laptop for my next. I just need someone
like HP to make it. I'm not spending $3K for the luxobooks.
> And for what...? 64-bits has very little demonstratable
> advantage today, and we are in no danger of missing out on
> important application features in the near future, since all
> 64-bit apps over the next few years will sport 32-bit
> versions.
In software engineering, there is a tipping point to where you
migrate your base development platform to the new stuff. Some
companies are ahead of the curve and more are behind the curve.
The ones that are ahead of the curve get the free systems from
the manufacturers before they are released and experiment with
them for a while and then make the decision to go with the new
hardware/software platform or not.
When they do that, less emphasis is placed on the older platform which then becomes legacy. Features for the legacy
platform are harder to justify and over time, engineers think
in terms of the new platform. And you tend to get more errors
in the older platform as it's not on engineers' minds WRT the special things that you have to remember about platforms.
Panorama already offers additional functionality in their
64-bit version compared to their 32-bit version. That's a
feature advantage today but it's also a performance advantage
were the software to support larger models using a 32-bit
platform as you have to stitch together an addressing scheme.
I see this in the 128-bit and higher encryption code in the
Mozilla code base. Right now, you have 128 biit operations that
are done with 32-bit registers so you have to stitch four 32-bit registers together and do all sorts of things to get that to work. You still have to do that with 64 bit registers but it's less than half the work.
A Sun engineer indicated that there's a compile-time switch
that will turn on 64-bit integer code for an RSA algorithm
that he has a 75% performance improvement on Solaris compared
to the 32-bit algorithm. So there are practical benefits to
64-bits. And of course the additional general purpose registers
and the additional multimedia registers.
I have some arguments about the way that Microsoft implemented
SIMD support in Windows x64 but the extra registers are a win
if you know how to use them.
I'm hoping for an HP DC 64-bit system in June. Of course my mailbox is always open if someone wants to send me a Yonah laptop to do performance testing on. I've never seen anyone
beat my BenchJS (JavaScript benchmark) using Firefox. Even
from folks with faster processors than I have. And yet it
still easily beats my fastest 32-bit implementations.
Market Data 
Markets 
AI
