On Apr 24, 2009, at 7:18 AM, n3evl wrote:
Woodrick, Ed wrote:
Now, if you had a regular radio, would you install a toaster in it?
What about using it as a remote control. Hopefully your answer is
no. So why do you make your SDR controller a Word Processor? A game
machine? A movie viewer?
I understand what you're getting at but, when it comes down to it,
this _isn't_ a regular radio! I fully expect the PC component of my
radio to be many other things as well.
Certainly. But what constitutes your "computer" may well change. More
below.
Yes, I am careful about what I allow on there but I have yet to be
persuaded that my _general purpose computer_ cannot remain exactly
that, including running PowerSDR!
This requires a change in mindset as to how things work. Sun
Microsystems has been saying it for years but it is true and correct,
that, "The Network is the Computer." It makes much more sense to
dedicate small computers to specific functions and then distribute
many of them no the network so that no single system carries all the
load. My home network currently has 9 or 10 machines scattered around
performing different tasks. Each is fairly small and more than half
have no rotating media (disk drives) because they are dedicated to
some function. One computer is dedicated to being a file server for
all the rest.
Personally, I am running a relatively small dedicated machine to power
my F5K. Right now this machine runs about 70% utilization because it
is running PSDR, VAC, com0com, ddutil, and fldigi. (I have stripped
most of the functions out of Windows to make this run efficiently.) I
expect this machine to get back a lot of capacity when we move to the
next generation of flex software and the radio display moves across
the network to my laptop, leaving just the DSP running on the
dedicated F5K machine.
To date, after obtaining a suitably fast machine I have had zero
problem running the radio and a variety of other applications and
have no intention of stripping this PC down to be a glorified radio
controller.
The proliferation of programs like cwskimmer are going to force you to
distribute. It will be possible to for the radio to multicast the I/Q
streams and to distribute the DSP so that you can effectively
demodulate and decode all the digital traffic on all the bands. It
*IS* going to happen but it isn't going to happen in a single machine.
As you say, performance is going to depend on a variety of
individual choices of hardware, software, and configuration, and
individual mileage will vary, but I do not accept it as a given that
the PC cannot continue to perform as a general purpose device.
Are you familiar with Moore's Law? It says that gate density, and
hence computing power, will double every two years. The amazing thing
about Moore's Law is that it has held up so well for over 50 years.
But we are seeing a shift in how that happens. In the past it was
accomplished by increasing the processing power of individual
processors by increasing gate density (more gates per unit of volume).
But as individual gates approach molecular size, that approach falters
as it becomes difficult to keep shrinking things. We have one more big
step and that is to quantum mechanical gates but after that, it will
be impossible to shrink the gate any further. At that point gate
density will remain fixed.
So the chip manufacturers have been continuing to adhere to Moore's
Law by moving to parallelism. The increase in performance is starting
to come from multiple processor cores. Since closely-coupled
processing using multiple cores is difficult, the power comes from the
splitting up the functions along the lines of logical interfaces and
loosely coupled processing. Once you do that it won't matter whether
the other part of your task is done in box A or box B around the
corner and across the network.
And then there is the final nail in coffin of the single central
processor: distributed computing is going to be a lot cheaper.
So you really do need to start thinking along these lines. The guys
working on the next generation of code recognize this and are working
to distribute the functions. You might want to consider embracing this
approach as, frankly, it is what is going to happen.
73 de Brian, WB6RQN/J79BPL
Brian Lloyd
Granite Bay Montessori School 9330 Sierra College Bl
brian AT gbmontessori DOT com Roseville, CA 95661
+1.916.367.2131 (voice) +1.791.912.8170 (fax)
PGP key ID: 12095C52A32A1B6C
PGP key fingerprint: 3B1D BA11 4913 3254 B6E0 CC09 1209 5C52 A32A 1B6C
_______________________________________________
FlexRadio Systems Mailing List
FlexRadio@flex-radio.biz
http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/
Knowledge Base: http://kc.flex-radio.com/ Homepage: http://www.flex-radio.com/