The "Good" Old Days - Complete Specs for DX-10 Operating System

[[e-mail]]

http://www.bitsavers.org/pdf/ti/990/dx10/0946250-9701B_DX10_Vol1_ConceptsAndFacilities_Apr81.pdf

. . .

DX-10 was writ mostly for the TI-990 mini-computers
in the mid 70s. There are a few pix in the PDF and
show people with clearly mid-70s styles.

The 990 series, and DX-10, were an early multi-user/
multi-tasking system aimed at the 'mini-computer'
segment. Most users would have had plain terminals or
graphics-capable serial terminals like from TexTronics.

The 990 series used the TMS-9900 chip and near variants.
This was an odd chip - kept the CPU registers out in
ordinary RAM and could switch quickly between different
sets of registers. At that time, the external RAM and
CPU kinda ran at the same speed so little was lost
putting the registers in RAM.

I remember fooling with this chip on a TI-99/4a home
computer (which, tragically, horribly under-used the
neat new 16-bit CPU). There were ASM commands for
dealing with the register, 'context', shifting. By
that means many users with their own space could
be implemented directly with the hardware.

Anyway, the link is the full manual for DX-10 and
also explains its symbiosis with the 990 hardware.

How it is now AIN'T how it always was.

[John Ames]

On Wed, 2 Oct 2024 03:18:23 -0400

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time, the
external RAM and CPU kinda ran at the same speed so little was lost
putting the registers in RAM.
I remember fooling with this chip on a TI-99/4a home computer (which,
tragically, horribly under-used the neat new 16-bit CPU). There were
ASM commands for dealing with the register, 'context', shifting. By
that means many users with their own space could be implemented
directly with the hardware.

I've always found the 9900 concept interesting, although its core
assumption about memory speed doesn't really hold up today; much of the
architecture was eventually reincarnated in TI's MSP430 series micro-
controllers, but they ditched the memory-resident register file. But
for the time, context-switching certainly didn't get any faster than
that; only three actual registers to save, but you still got a
comfortably PDP-11ish environment from the programmer's perspective.

But yes, it's astonishing just how bad the TI-99 design was; a cascade
of bad decisions that turned what could've been a real contender in the
home-computer wars into an absolute joke. One of my oneathesedays
projects that I've toyed with for years (though never made any headway
on) is to roll up a homebrew system that does right by the concept...

How it is now AIN'T how it always was.

And may it never be forgotten ;)

[John Ames]

(It's also, I gather, another case of the truly bizarre management
philosophy at big tech companies in the '60s - '70s where different
departments of the same business were forced to Thunderdome each other
for the amusement of the C-suite, and everyone was then Very Astonished
to find that their company was bleeding out from a profusion of self-
inflicted stab wounds...)

[rbowman]

Post by John Ames
I've always found the 9900 concept interesting, although its core
assumption about memory speed doesn't really hold up today; much of the
architecture was eventually reincarnated in TI's MSP430 series micro-
controllers, but they ditched the memory-resident register file. But for
the time, context-switching certainly didn't get any faster than that;
only three actual registers to save, but you still got a comfortably
PDP-11ish environment from the programmer's perspective.

I worked on one project that used the 9900. Its claim to fame is TI had a
rad hard version.

[John Ames]

On 2 Oct 2024 16:35:00 GMT

Post by rbowman
I worked on one project that used the 9900. Its claim to fame is TI
had a rad hard version.

Interesting - did that see use in nuclear power? Unlike the RCA 1802,
it seems way over-featured for use in space probes...

[rbowman]

Post by John Ames

Post by rbowman
I worked on one project that used the 9900. Its claim to fame is TI had
a rad hard version.

Interesting - did that see use in nuclear power? Unlike the RCA 1802,
it seems way over-featured for use in space probes...

No, the application was military aircraft. The Soviets had the real answer
-- vacuum tubes. The lesson is never learned leading to littoral combat
ships that can be taken out by a jihadi in a Zodiac.

I think it's urban legend but there is the story about developing a ball
point pen that works in zero gs. The Soviets used pencils.

[John Ames]

On 2 Oct 2024 22:28:24 GMT

Post by rbowman
No, the application was military aircraft. The Soviets had the real
answer -- vacuum tubes. The lesson is never learned leading to
littoral combat ships that can be taken out by a jihadi in a Zodiac.
I think it's urban legend but there is the story about developing a
ball point pen that works in zero gs. The Soviets used pencils.

Yeah, that tracks XD

[Lars Poulsen]

Post by rbowman

Post by John Ames

Post by rbowman
I worked on one project that used the 9900. Its claim to fame is TI had
a rad hard version.

Interesting - did that see use in nuclear power? Unlike the RCA 1802,
it seems way over-featured for use in space probes...

No, the application was military aircraft. The Soviets had the real answer
-- vacuum tubes. The lesson is never learned leading to littoral combat
ships that can be taken out by a jihadi in a Zodiac.
I think it's urban legend but there is the story about developing a ball
point pen that works in zero gs. The Soviets used pencils.

A pencil sharpener in zero-G is a huge problem.

[Charlie Gibbs]

Post by Lars Poulsen

Post by rbowman

Post by John Ames

Post by rbowman
I worked on one project that used the 9900. Its claim to fame is TI had
a rad hard version.

Interesting - did that see use in nuclear power? Unlike the RCA 1802,
it seems way over-featured for use in space probes...

No, the application was military aircraft. The Soviets had the real answer
-- vacuum tubes. The lesson is never learned leading to littoral combat
ships that can be taken out by a jihadi in a Zodiac.
I think it's urban legend but there is the story about developing a ball
point pen that works in zero gs. The Soviets used pencils.

A pencil sharpener in zero-G is a huge problem.

The story I heard was that graphite dust would short out microelectronics.

[Peter Flass]

Post by rbowman

Post by John Ames
I've always found the 9900 concept interesting, although its core
assumption about memory speed doesn't really hold up today; much of the
architecture was eventually reincarnated in TI's MSP430 series micro-
controllers, but they ditched the memory-resident register file. But for
the time, context-switching certainly didn't get any faster than that;
only three actual registers to save, but you still got a comfortably
PDP-11ish environment from the programmer's perspective.

I worked on one project that used the 9900. Its claim to fame is TI had a
rad hard version.

Skidmore College had one.

[rbowman]

Post by Peter Flass

Post by rbowman

Post by John Ames
I've always found the 9900 concept interesting, although its core
assumption about memory speed doesn't really hold up today; much of
the architecture was eventually reincarnated in TI's MSP430 series
micro- controllers, but they ditched the memory-resident register
file. But for the time, context-switching certainly didn't get any
faster than that; only three actual registers to save, but you still
got a comfortably PDP-11ish environment from the programmer's
perspective.

I worked on one project that used the 9900. Its claim to fame is TI had
a rad hard version.

Skidmore College had one.

Figures... I went to RPI years before the 9900 was even a dream. Let's
just say the girls from Skidmore, Vassar, and Bennington weren't the most
approachable for nerdy engineers. I hadn't thought about Skidmore in
decades.

[Lawrence D'Oliveiro]

Not too much later, we replaced it with a
recorder from Radio Shack (a close relative of the one they sold for use
with their TRS-80s), and that one worked like a champ.

Those cassette interfaces were purpose-built for data use, apparently.
That’s why they worked so well.

[Rich]

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

[The Natural Philosopher]

Post by Rich

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

The 6502 was an excellent little beast. I never programmed one myself,
but I know people who did, and my friend who worked on the original
Acorn machines said it was pretty fast compared with a Z80 etc

Working with that is what led to the Acorn Risc Machine, which we all
know and love as the ARM architecture

They couldn't afford to make a big chip, so they worked on an extremely
small one, and gave it an instruction set that was minimal, but powerful
. Just like the 6502.

[Lawrence D'Oliveiro]

Post by The Natural Philosopher
The 6502 was an excellent little beast. I never programmed one myself,
but I know people who did, and my friend who worked on the original
Acorn machines said it was pretty fast compared with a Z80 etc

The 6502 got its speed by restricting itself to 8-bit addressing in a lot
of places. E.g. the stack was limited to 256 bytes and had to be located
in memory page 1. The Z80, on the other hand, allowed for full 16-bit
addresses in a lot more places.

[Peter Flass]

Post by Rich

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

Back when microprocessors were just getting started I researched specs for
8-bit chips - instruction set, I/O, etc. The only one I thought was worth a
bucket of warm spit was the 6800.

[rbowman]

Post by Peter Flass
Back when microprocessors were just getting started I researched specs
for 8-bit chips - instruction set, I/O, etc. The only one I thought was
worth a bucket of warm spit was the 6800.

The problem was Motorola thought highly of the 6800 and didn't want to
hear the thoughts of the breakaway team that designed the 6502 to be
cheaper and faster.

[Scott Alfter]

Post by Rich

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

...and the 65816 (the 16-bit successor to the 6502 that was used in the
Apple IIGS) let you move what was now called the "direct page" anywhere in
the first 64K of memory. A new register for the purpose was added, along
with another one that allowed the stack to also be put anywhere in the first
64K (vs. having it locked to page 1).

[Rich]

Post by Scott Alfter

Post by Rich

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

...and the 65816 (the 16-bit successor to the 6502 that was used in the
Apple IIGS) let you move what was now called the "direct page" anywhere in
the first 64K of memory. A new register for the purpose was added, along
with another one that allowed the stack to also be put anywhere in the first
64K (vs. having it locked to page 1).

Nice, I did not know that about the 65816.

I do wonder why restrict it to only first 64K?

[Scott Alfter]

Post by Rich

Post by Scott Alfter

Post by Rich

The 990 series used the TMS-9900 chip and near variants. This was an
odd chip - kept the CPU registers out in ordinary RAM and could
switch quickly between different sets of registers. At that time,
the external RAM and CPU kinda ran at the same speed so little was
lost putting the registers in RAM.

The 6502 did something similar. It wasn't as far down the path as the
TI chip, but page zero (first 256 bytes of ram) acted a lot like an
'extended register file'. There were even addressing modes that used
two consecutive bytes of "page zero" as a 16bit pointer into the rest
of the RAM one's system had installed,

...and the 65816 (the 16-bit successor to the 6502 that was used in the
Apple IIGS) let you move what was now called the "direct page" anywhere in
the first 64K of memory. A new register for the purpose was added, along
with another one that allowed the stack to also be put anywhere in the first
64K (vs. having it locked to page 1).

Nice, I did not know that about the 65816.
I do wonder why restrict it to only first 64K?

https://softpixel.com/~cwright/sianse/docs/65816NFO.HTM#3.00

The D (direct-page pointer) and S (stack pointer) registers are 16 bits wide
in native mode. This allows the direct page and stack to start anywhere
within the first 64K, not just on page (256-byte) boundaries. The 16-bit S
also means you can have a stack larger than 256 bytes, though stack
operations could potentially stomp on the direct page.

In (8-bit) emulation mode, the high byte of S is masked to 0x01, so the
stack always occupies the space from 0x000100 to 0x0001FF. D remains 16
bits wide, but zero-page addressing wraps around within the selected page,
so that if (for instance), D is set to 0x0080, "LDA $00" will read physical
address 0x000080 and "LDA $7F" will read physical address 0x0000FF, but "LDA
$80" will wrap around to read physical address 0x000000.