[PLUG] 64-bit programming

briand at aracnet.com briand at aracnet.com
Mon Jun 17 06:44:52 UTC 2002


>>>>> "Steve" == Steve Bonds <1s7k8uhcd001 at sneakemail.com> writes:

  Steve> Two big reasons:
  Steve> 1) Memory space (no 4GB limit)

Not really true.  For execution there is no reason why the memory
access needs to be limited to the data bus width.  I could easily make
my memory accesses take place on a 128 bit bus and fetch only 16, 32,
64, etc, at a time.  In theory the design might be a little ugly since
I now need to do 128 bit math for the program counter but it could be
done.  A 64 bit wide instruction set is pretty much a waste.

  Steve> 2) Single-threaded performance (SMP is great for
  Steve> multithreaded apps, but not everything can be multithreaded)

 Single threaded performance definitely benefits from moving the 64
 bit chunks around.  An interesting class of programs are the math
 intensive, or CAD oriented programs.  Heavy duty math programs are
 almost ALWAYS very parellizable (is that a word ?).  They typically
 do a lot of matrix math.

The future is in parallelization.  For math applications there is no
doubt in my mind that people should be focusing on lots of little
processes working together - and to a large extent they are - hence
all the hoopla about clusters.

As for the "lay person" bigger and better CPU's are already a dead
end.  Who the hell needs a 1.8GHz P4 for word processing, or instant
messaging ?  Even BillG is starting to have trouble generating enough
code bloat to slow down modern processors.  The only reason I've ever
upgraded my system was to crunch FP faster.  As far as usability and
day-to-day use I could get by with a K2-233.  The boost in HD
performance has really made the difference for everyday use.

More performance for the 3D graphics world still has lots of life left
in it.  Visualization is very BW and computationally expensive.
Again, graphics are also very parallelizable.

Bottom line is that Intel and AMD need to put out bigger and better
chips strictly for profit margins, and they desperately need to make
people think that they need them - but they don't.


Brian


  Steve>   -- Steve

  Steve> On Sat, 15 Jun 2002, Carla Schroder pluglist at bratgrrl.com XXXXXXXXXXXXXXXXXXX wrote:

  >> Any thoughts on programming for a 64-bit cpu like the Itanium, other than it 
  >> sounds like a speech impediment? Are there practical reasons? Kylix is 
  >> strictly IA32, and may someday support 64-bit cpus. Then again, maybe not. 
  >> gcc of course has IA64 and Alpha support. I'm wondering does it really matter 
  >> in these days of distributed computing and low-cost clusters and SMP, why 
  >> complicate things by introducing yet another complication? Just string
  Steve> a herd 
  >> a them IA32s together.
  >> 
  >> Thanks,
  >> Carla
  >> 
  >> _______________________________________________
  >> PLUG mailing list
  >> PLUG at lists.pdxlinux.org
  >> http://lists.pdxlinux.org/mailman/listinfo/plug
  >> 




  Steve> _______________________________________________
  Steve> PLUG mailing list
  Steve> PLUG at lists.pdxlinux.org
  Steve> http://lists.pdxlinux.org/mailman/listinfo/plug





More information about the PLUG mailing list