Re: [uCsimm] cs8900 & wait-states

From: Bob Vincent (bvincent@bbn.com)
Date: Tue Dec 12 2000 - 18:06:32 EST


Argument for bigendian (intel) net order:
Numbers can be shifted in and added as they're received (works best for
bigendian bytes, too!)

I believe the whole controversy began when intel decided they liked the ms
byte in the low (even) byte of a word. That is "natural" (see above) if you
are doing 16 bit arithmetic on an 8 bit bus.

The rest of the world (pretty much) believed the ls byte should live in the
low (even) byte. Just happens to be a useful net order, because you always
know what the whole number is when it stops getting sent, no matter how
many bytes you sent. Unfortunately, net thinking came after hardware
optimization went out the window.

Only my opinion, based on personal experience.

At 03:17 PM 12/12/2000 -0500, you wrote:
>Florian:
>
>You'll probably get way too many replies for this one, and I'm
>probably not the first.
>
>Here's a question: if you're transmitting a multiple-byte
>value over a network, which would you send first? The MSByte
>or the LSByte? There are good arguments for either order.
>
>Your argument for big-endian ordering is compelling: you can read
>bytes the way they appear in memory.
>
>But mightn't it make sense that the least significant byte of a
>multi-byte word appear in the least significant address? That's
>the little-endian argument.
>
>Both viewpoints are valid. They're just different.
>
>- Rob
>
>
>"Florian G. Pflug" wrote:
> >
> > On Thu, Dec 07, 2000 at 02:10:09PM +0100, sticke_m wrote:
> > > There is no reason than just historical of putting the low
> > > or the high byte of a word at first in memory.
> > > For motorola you can write $12345678 in memory for a $12345678
> > > and for intel you write $78563412. It's just a historical reason.
> > Hi
> >
> > Just for interesset...
> >
> > Does anyone know if it ever had any advantage to do it one or the other way
> > round?
> >
> > The big-endian format seems much more logical to me (after all,
> $12345678 is
> > $12345678 in memory with this).
> >
> > So there had to be some reason, why someone ever hat the idea to swap the
> > bytes and store it as $87654321....
> >
> > Or, maybe this was just some mistake in an early chip/board layout? ;-))))
> > Just like once, long long time ago people actually had an "unmount" command
> > - until someone forgot the "n" in an BSD-Release...
> >
> > greetings, Florian Pflug
> > This message resent by the ucsimm@uclinux.com list server
> http://www.uClinux.com/
>This message resent by the ucsimm@uclinux.com list server
>http://www.uClinux.com/

Bv)

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 - 00:01:39 EST