From Newsgroup: comp.arch

On Sun, 24 May 2026 19:00:30 -0000 (UTC), quadi wrote:

The STRETCH had a 64-bit word, and so it anticipated one feature of
the System/360. The word "byte" was coined for it, and it had bit
addressing.

The original meaning of rCLbyterCY was not rCLminimum directly-addressable
unit of memoryrCY, as at present, but rCLvariable-length bitfieldrCY.

In the 1401, each character had an additional bit which served as a
word mark associated with it. The 705 didn't do that. But in both of
those computers, the address of a number was the address of the
character containing its least significant digit. Just like it would
be in a little- endian machine. But these machines were big-endian,
and so during addition, the computer counted backwards to find
successive digits to add and propagate the previous carry to.

Addition and subtraction started from the lowest-addressed digit, and
proceeded upward from there. This allowed for arithmetic (well,
addition and subtraction, anyway) on arbitrary-length numbers.
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: comp.arch

According to quadi <quadibloc@ca.invalid>:

The System/360 was intended to replace both the 7090 series and the 1401 >series, as well as other IBM machines.

Right. I would say that it was more important to replace the 1401 and other commercial machines, since IBM sold a lot more of them than they did of the high
end 709x. All of the microcoded 360 models, 25, 30, 40, 50, 65, and 85, had emulators for the older machines. IBM only provided emulators for models where the emulation would run faster than the original machine, so there were 1401 emulators for the /25 and /30 but you needed a /65 to emulate a 7090.

So the use of SS instructions for BCD arithmetic arose from trying to have this portion of the instruction set as similar to the previous commercial
IBM machines as possible. But the 360 didn't indicate the end of a decimal integer by a word mark or by a special character; the SS instructions included length fields.

Numbers were addressed by their most significant digits; the contents of
the length field could be added to that address if required.

They didn't want word mark bits because they'd have made no sense in the halfword and fullword scientific instructions. The field lengths were
hard coded into the second byte of SS instructions, but you could fake
variable length operands with the EX (execute) instruction. It took
a register operand and OR'ed the low byte of the register with the
second byte of the target instruction before executing, so you could put
a zero length in the instruction and the actual length in the
register. It was a cute hack.
--
Regards,
John Levine, johnl@taugh.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly
--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: comp.arch

On Sun, 24 May 2026 22:49:58 +0000, Lawrence DrCOOliveiro wrote:

On Sun, 24 May 2026 19:00:30 -0000 (UTC), quadi wrote:

The STRETCH had a 64-bit word, and so it anticipated one feature of the
System/360. The word "byte" was coined for it, and it had bit
addressing.

The original meaning of rCLbyterCY was not rCLminimum directly-addressable unit of memoryrCY, as at present, but rCLvariable-length bitfieldrCY.

Yes, that's correct. On the STRETCH, not only were bytes bit-addressed,
but they could cross the boundaries of the system's 64-bit words with no problem.

The PDP-10, on the other hand, also had variable-length bytes, but they
could not cross word boundaries. Even so, it was the machine that came
closest to STRETCH in retaining the original meaning of a byte.

John Savard
--- Synchronet 3.22a-Linux NewsLink 1.2