On 2013 Apr 9, at 5:41 PM, Liam Proven wrote:
On 10 April 2013 01:00, Kelly Fergason <kfergason
at gmail.com> wrote:
TI 99/4a. Cliff Click gave up in disgust trying to program this
thing. nuf said.
You do have a point. Interesting architecture & so on, but...
registers in external DRAM? Really?
The "registers in external (D)RAM" was not a problem, that was a
characteristic (and feature) of the predecessor TI 990 minicomputer
architecture. It meant the instruction set had the efficiency of a
very orthogonal set of 16 registers to work with, but didn't have to
save & restore those registers across function call & return, you
just changed the workspace pointer and had a whole new set of
registers. This was great for modern stack-based languages and for
process context-switching (if you were doing that). It was far nicer
than a two-accumulator architecture such as the 6502 or the mish-mash
register set of the 8080/Z80. I worked with the 990 mini and 9900
microprocs at the instruction level and, in that regard (and as much
as I remember), I liked the arch. more than the PDP-11.
I didn't directly work with the TI-99/4 but from what I read the
problem with it was it took what was a lovely 16-bit microprocessor
and embedded it in a crippled memory/support environment. Apparently
only a few hundred words of RAM were directly on the processor bus -
that mem being intended for the workspace/registers (and presumably
this actually was static ram, not DRAM) - while the majority of RAM
was accessed through a slower, secondary, 8-bit bus.
They were appallingly slow. ...