Re: [uCsimm] net traffic filtering

From: Tom Walsh (tom@cyberiansoftware.com)
Date: Tue Jan 16 2001 - 21:05:29 EST


Andrew Kohlsmith wrote:
>
> > I tend to agree with Daniel, the EZ is not a performance platform, it
>
> Hmmm... I was wondering if you could elaborate on the various 68000
> implementations Motorola has out... i.e. What are the
> strengths/weaknesses/relative prices to Coldfire, Dragonball EZ/VZ/etc and
> so on and which are currently supported by uClinux...
>
> I've looked around the website (and various affilliated websites) and the
> information seems to be only in bits and pieces. Someone such as yourself
> could go a long way in clearing all this up for me (and I'm sure for other
> lurkers too).
>

Well, if you insist on using a 68K processor, why not look into the
MC68EN360? It is 8 MIPS, has native 802 (ethernet) protocol implemented
which can be used on one SIO channel. There is some active work going
on at Lineo on this platform. It is not in CVS, but I'm sure if you ask
they will give you the current patches. As to strengths / weaknesses,
you have to examine those yourself, designing / choosing a hardware
platform is *always* a series of trade-offs as there is no such thing as
the "perfect solution". I did get a set of patches myself, but have not
placed them in my CVS.

The distinct advantage that the VZ has over the EZ (aside from MIPS) is
that the VZ will interface to SDRAM (up to 32 Meg of it), this is pretty
persuasive in itself! The VZ also has an additional 16bit timer, plus
an additional serial port...

Are you designing your own, or looking to purchase an off-the-shelf
solution? As far as cost in concerned, that is not in my equation,
development cost is what my customers are sensitive too (low volume OEMs
that do a lot of customization). A lot of the really nice processors
are going BGA (Ball Grid Array), that technology really raises the
development bar (and costs).

As you can see from my website, 'http://www.openhardware.net', I have
been working with the MC68EZ328 processor. For the Ez328Lcd board, and
its' intended application, it is *adequate*, not a perfomance monster by
a long shot! In order to get decent performance when drawing iconic
images on the QVGA Lcd, I wrote my own set of graphics routines. You
simply cannot use more advanced stuff like Micro-X, etc, and get away
from the sluggish drawing of images on the screen. With my own
routines, I have more of a "snapshot" effect of changing images.

This is why I would question the use of the EZ for a packet sniffer, it
all depends on what you expect to be able to do with it. These
controllers are no where near the performance of a 386SX-33! Try it,
dig up an old 486SX motherboard, shove 8 Meg of DRAM on it, drop the
clock WAY back (as low as you can go, 500 kHz?) and try to do some
serious work with the board. Keep in mind, the 68K processors are not
as fast as the X86 processors are at executing a similar set of
instructions. Clock per clock, the 68K processors run at a magnitude
of, perhaps, 8..10 in performance *reduction* when compared to an
equivlant X86 at the same rate. If you look carefully at the clock
cycles needed to perform various operations (Mulitply, Subtract, etc),
you will see a marked difference in how the x86 uses less clock cycles
vs. the 68K.

When comparing X86 to ARM, note that the X86 uses "macro instructions",
lots of specialized opcodes with added pipeling to get the performance
up. ARM uses many smaller efficient instructions running at a higher
clock rate to get their performance up. It is sad, but Motorola has
done nothing with the 68010 core, the higher numbered cores (68020, 030,
...) are successive improvements, but we cannot use those for embedded
work... Too much amperage to run them! The 68302 core had been
developed for the automotive market for in-vehicle engine control. When
the demand ran out in Detroit, Motorola found itself with an embedded
systems market consuming the processors and they kept the line going as
demand was high enough to justify it. IMHO, they stumbled into Palm
using it in their PDA... When Palm stops using the 68EZ328 in their PDA,
watch out(!), it will probably go EOL (End Of Life).

IMHO, it is an Apples & Oranges comparison. The DragonBall is low
power, but low performance. Arm is also low power, but high
performance. X86 is just out of sight, the lessor 386 embedded cores
are being EOL'ed (End of Life) and they don't seem to understand that
the embedded market doesn't need the performance of the '586 to do a lot
of the jobs.

Each of the processor types present you with different challanges: 68K
is simple to implement, but low MIPS. ARM is high performance, but
there is a learning curve on implementation + the RAM has an odd mapping
scheme that you have to study & master. The X86 is nice, but AMD &
INTEL seem to be moving rapidly away from their low-end offerings (386)
and seem to believe that everyone will follow. Yes, there is Nat' semi's
offering of an SIOC, but how long will they stay in the market (their
past track record in the embedded market is dismal)?

Personally? I am moving very quickly into ARM. I have spent the past
several weeks considering various chipsets for use in two new designs.
One design MUST remain an X86 platform as the data is odd word aligned
(only X86 can successfully deal with that condition). The other design
MUST be an ARM processor, X86 is out as the only solution looks to be
going EOL soon and the darned thing has no integrated DRAM controller
(Intel386EXTC). The ARM design will require about 20-40 MIPS, have an
MMU, and DRAM controller, and run at fairly low power consumption; my
choice looks to be the Cirrus Logic 7212.

Okay, I have "run off at the fingers" long enough. BTW, I am not an
expert, just a hacker. The opinions expressed are my own and can be
easily challanged by others more knowledgable ... :-)

Regards,

TomW

-- 
Tom Walsh - WN3L - Embedded Systems Consultant
'www.openhardware.net', 'www.cyberiansoftware.com'
"Windows? No thanks, I have work to do..."
This message resent by the ucsimm@uclinux.com list server http://www.uClinux.com/



This archive was generated by hypermail 2b30 : Sun Apr 07 2002 - 00:01:40 EST