Wirt Atmar wrote:
>
> John Burke writes:
>
> > You left out Control Data. I spent many, many hours on a 6600. BTW, a
> >  byte was 6 bits ;-)
>
> One of my pet peeves (I'm just like Stan in that regard) is these constant
> Orwellian re-writes of history, even if accidental. In that regard, the
> definition of the "byte" is now completely fouled from its original
> definition.
>
> Originally, a "word" size was any length you specified -- and wholly dependent
> on the binary resolution you wanted. It had nothing to do with hardware. It
> was simply arithmetic in its definition.
>
> Clearly, however, very few early processors could process a whole word in one
> GULP. Rather, they had to do it BITE size pieces, in a series of staggered
> steps. If you had an 8-bit processor, your "byte" (notice the clever
> respelling) size was 8 bits. If your processor was 16-bits wide, then your
> "bytes" were 16-bits. And so on.
>
> And in continuance of the joke, a "nibble" was defined as half a byte.
>
> But, somehow, over time, a byte became stuck at 8 bits, a nibble at 4, and an
> HP3000 word at 16 bits -- which became redefined with the advent of the RISC
> processors to be 32 bits, and the old 16-bit words demoted to "half-word"
> status, just to further confuse things.
>
> But in actual fact, an R4 number in IMAGE has a word size of 64-bits and a
> RISC machine has a byte size of 32-bits and a nibble size of 16 bits -- or at
> least, it should.

"Somehow over time"?.  I can shed some light on that.  IBM (the de facto
standard
25 years ago) came out with the 360.  Before that IBM had "commercial"
processors, the 1401, 1410, of small to medium size that "were"
character machines,
i. e., the operands were of variable length.  The characters had six
data bits, plus
a "word mark" to represent the end of an operand (also a parity bit, but
that need not
concern us here).  IBM's 7000 series, 7040, 7070, 7090 and some later
models of
these were word machine - they did not use a word mark, but had fixed
length words.

 In the 360 series, IBM went to an 8 bit "character".  But the 360
had other ways of of using these units, and IBM called them "bytes"
rather
than characters.  IBM also called its disc drives "disk" drives.  After
all,
the marketing people had to justify their existence somehow.  Anyway,
now we
are stuck with this terminology.  The 3000 borrowed these data
structures.

Side note.  You can tell that SOME people at HP weren't happy with the
IBM
structures.  the writers of SPL did not see fit to include the packed
decimal instructions( an IBM invention) as part of SPL, even though the
classic
3000's had packed decimal machine instructions.  An SPL programmer had
to
descend to an "assemble" statement to use them.

Enough history?

Nick D.