On 2024-06-13 12:06 p.m., Paul Koning via cctalk wrote:
On Jun 13, 2024, at 2:00 PM, Chuck Guzis via cctalk <cctalk@classiccmp.org>
wrote:
On 6/13/24 10:32, Paul Koning via cctalk wrote:
Huh? There is no direct connection between word length, register count, and
pipeline length.
Indeed. There are architectures with NO user-addressable registers.
Some have memory-mapped registers, where a "register number" is merely
shorthand for a memory address (i.e. implicit multiplier and base)
--Chuck
There are of course also machines that appear to have registers (in the sense that the
instruction set refers to them) but the implementation is a chunk of memory. The PDP-6
is one (and early PDP-10s without the "fast registers" option). The Philips
PR-8000 may be one as well; the ISA has 8 registers, times 8 because it has a separate
set per interrupt priority level, but there is a variant of the store instruction that
stores to any of these as if it were memory. I'm not actually sure if it was implemented
that way; 64 registers 24 bits wide would be a substantial cost and bulk in a mid-1960s
machine.
paul
Index registers like on, the IBM 1130 worked that way.
The PDP-5 I think had the fewest internal registers.
(Never heard of Nuclear plan fail with PDP-5).
B.
