[NLRS] 439MHZ Radar

Dr. Gerald N. Johnson, electrical engineer [email protected]
Fri, 11 Apr 2003 21:44:23 -0500


For sure wide words and multiple busses improve the throughput, but the
basic clock speed and memory addressing helps. I gave up on main frames
nearly 30 years ago, every time I wanted to run a complex piece of
software the OS had been updated and the object file wasn't compatible
with my previous compile which made my computing costs skyrocket because
compiles cost several times what production runs did. In those days, I
don't know what the CPU speed was but RAM wasn't as big as 32 MB and
runs cost more per hour that we can buy a PC for today.

There are those that claim a gaggle of networked PCs running Linux with
software properly compiled for such multiple processors can process
large scientific tasks very competitively compared to large main frames.
My software runs too sequentially to be so split though.

Most generations of Intel/AMD '86 family do make significant
improvements in the number of clock cycles per operation. All are
dependent on programs and data fitting within the on-chip cache.
Swapping tasks is a performance killer. Programs written in "convenient"
programming languages instead of C or Assembler tend to be inefficient
enough to make each faster generation of CPU run not much faster than
the previous CPU. John Dvorack's comment years ago that Windoze was to
make the new '286 run no faster than the 8088 still seems to be true.

RISC architectures try to proof their adequacy by claiming that
compilers don't use most of the CISC instruction set. I guess that
points at compiler creators neglecting to use more of the benefits of
the CISC instruction set. But then I remember finding that most of the
Z80 CISC instruction set extensions beyond the 8080 used more clock
cycles than doing the same task with 8080 instructions.

I did do some assembler in the 6502 also, sort of a RISC. I found it
frustrating compared to the "CISC" of the Z80. And it wasn't hard core
RISC.

I depend little on windoze, most of my software is created for OS/2, the
business OS of European PCs and some for Linux. Both OS that make
windoze look like swiss cheese.

The PC structure lacks the I/O expansion capability of a mini or main
frame for remote I/O other than TCP/IP.

Years ago I concluded that in my one man operation selecting a CPU chip
had to be based on a little hype and a lot of hope because truly
evaluating the 8 bit chips of that era seemed to need a couple man years
each, an impossible task for projects expected to be completed in a few
months. So I chose the Z80 for those projects and it worked out. Today
learning enough about a CPU chip to do a good evaluation can be tougher
because the chip, its associated hardware, and the instruction set are
so much more complex. Its very hard to go away from PC based hardware on
cost alone. And I program almost always in C for its capability and
portability.

73, Jerry, K0CQ

-- 
Entire content copyright Dr. Gerald N. Johnson, electrical engineer.
Reproduction by permission only.