Then there was the 1802 RCA COSMAC CPU. A fellow CPUist, back then,
had one of those and we used to compete writing code (hand assembly) to
see who could do job X with less bytes of code. Man! That was a
primitive CPU with 16 16bit registers and not much else. No predefined
programme execution pointer, no 16 bit operations, even though all 16
bit operations, 8 bit working register. It was like writing microcode
but after you wrote standard call and return routines (subroutine
calls) the thing really kept up with any other 8 bit CPU.

I thoroughly enjoyed bit twiddling and miss the speaker buzzes and
light flashing.

One job we did was both wrote control software for a line printer
mechanism we picked up surplus. Timing counting and interrupt handling
was all great innovation using a massive 1024 bytes to get real
UPPERCASE printouts for hardcopy. We saved at least $3-4K each for a
dot matrix printer and had a lot of fun! The RCA 1802, with it's
primitive instruction set kept right up with anything I could do on a
Mot 6800 in byte count and speed.

Hadn't heard of a Nat SemiCon CPU before. Sounds like it was a
"minicomputer" back then.

-----------------
Larry Blanchard wrote:
I always thought the Nat'l Semi 16032 and its relatives had the best
instruction set of the micro chips, but my real favorite was a
mainframe from the '60s, the GE4xx series. Not a lot about it on the
web but a little at:

-------------
On Tue, 18 Oct 2011 00:57:21 -0400, m II wrote:

Yup, Motorola kept reducing the instruction sets down to 27? on the
68000 while Intel kept bragging about the millions of instructions on
it's messy internal architecture. People writing code for Intel hated
it!

Then came along the Z-80 with it's block memory move in one
instruction. Trouble was, that a 6809 with code for a memory move
loop could move the block faster and with less setup instructions
and time.

We all learned the long and hard way.