On 10/17/2011 11:21 AM, Larry Blanchard wrote:
....

I remember writing code for NASA way back when that filled a DMA buffer
with a pattern and then checked the pattern to see when I could flip the
pointers and start refilling the buffer. Or replacing one instruction
with an equivalent one because the second was a hair faster. Things were
different then.

Indeed they were.

I remember the "unlimited" expansion of capability when the systems I
was working on at one time went from 1- to 2(!!!!)-MHz clock cycle time
(w/ the incredibly rich instruction set/addressing modes) of a M6809E.

W/ the indirect addressing mode it made an almost perfect silicon
implementation of a Forth interpreter engine w/ the ";" next operator
implementable in, iirc, 2 instruction cycles as compared to 5 or 7 on
6502 or some other similar at the time.

--