You remember a list of Pintaske, mentioning the exploits of the
GA144 crew? Nothing of interest there. (Not the fault of Pintaske).
I had some hopes that now the patent troll Patriotic is no
longer involved, things would light up, but apparently not.

Then I remembered an interesting talk, I came across the other
day in youtube. So I'd share it with you.
Impossible to find back! No plausible keywords help.
I literally have to wade to all youtube films I viewed over the
last few weeks. Then I found it:
https://www.youtube.com/watch?v=xYWa2C2_7H0
hidden in the non-assuming title "SVFIG 17Nov2018 Afternoon"
that goes on rambling for more than 3 hours.
If you have patience than you find it at 1:09:
A real application! A heart rate monitor!
I concede that heart rate monitors comes free with smart watches
nowadays, but it is really the first time I came across
a description that some one has actually done some with the GA144.

(I discount rumors that Jeff Fox apparently did something with
hearing aids. )

It is interesting that the claim discussed on a 50+ thread
in electronics.design that apparently the GA144 can use
a 32 Khz watch crystal hanging off one of its ports
without any passive or active components needed,
is true.

Groetjes Albert
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 2:20:25 PM UTC+2, Hans Bezemer wrote:

On Monday, September 12, 2022 at 1:01:23 PM UTC+2, none albert wrote:

A real application! A heart rate monitor!

I don't know if you consider this "an application", but I found this: https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

.. and this one: https://dl.acm.org/doi/10.1145/2666356.2594339

Hans Bezemer

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 1:01:23 PM UTC+2, none albert wrote:

A real application! A heart rate monitor!

I don't know if you consider this "an application", but I found this: https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

Hans Bezemer

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 8:22:15 AM UTC-4, the.bee...@gmail.com wrote:

On Monday, September 12, 2022 at 2:20:25 PM UTC+2, Hans Bezemer wrote:

On Monday, September 12, 2022 at 1:01:23 PM UTC+2, none albert wrote:

A real application! A heart rate monitor!

I don't know if you consider this "an application", but I found this: https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

.. and this one: https://dl.acm.org/doi/10.1145/2666356.2594339

I like that they describe the, "the GreenArrays GA144, an extremely minimalist low-power spatial architecture that requires partitioning the program into fragments of no more than 256 instructions and 64 words of data."

The 64 words are program (256 instructions) and data. This is what makes the partitioning so difficult, the pieces have to be so very tiny. I suppose "spatial architecture" is a new term. I've never heard that before.

I believe an Ethernet interface was done using some number of nodes. The hard part seems to have been figuring out how to partition the design.

Has anyone seen much work on the GA144 in signal processing? That is an application area where data flow architectures have been used before. The GA144 would work very well as a data flow machine. Again, the hard part if designing the partitions.

The hearing aid application Jeff Fox talked about, would have been signal processing. The claim was that they had implemented a very unique algorithm on a TMS320C6xxx processor, and it took TWO of them burning some hefty amount of power. They managed
to stuff it into a single GA144 (which is feasible if the bottleneck was raw CPU cycles) at an unspecified power level. No one knows what happened commercially with this project. You have to assume it didn't become a product, or GA would have produced
a second batch of processors.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 9:01:23 PM UTC+10, none albert wrote:

You remember a list of Pintaske, mentioning the exploits of the
GA144 crew? Nothing of interest there. (Not the fault of Pintaske).
I had some hopes that now the patent troll Patriotic is no
longer involved, things would light up, but apparently not.

Then I remembered an interesting talk, I came across the other
day in youtube. So I'd share it with you.
Impossible to find back! No plausible keywords help.
I literally have to wade to all youtube films I viewed over the
last few weeks. Then I found it:
https://www.youtube.com/watch?v=xYWa2C2_7H0
hidden in the non-assuming title "SVFIG 17Nov2018 Afternoon"
that goes on rambling for more than 3 hours.
If you have patience than you find it at 1:09:
A real application! A heart rate monitor!
I concede that heart rate monitors comes free with smart watches
nowadays, but it is really the first time I came across
a description that some one has actually done some with the GA144.

(I discount rumors that Jeff Fox apparently did something with
hearing aids. )

It is interesting that the claim discussed on a 50+ thread
in electronics.design that apparently the GA144 can use
a 32 Khz watch crystal hanging off one of its ports
without any passive or active components needed,
is true.

Groetjes Albert
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

I think if GA had got large commercial sales, they would use a new verified batch,
rather than the old hobbyst developer batch. You can simply cut down you array dimensions to fit this way, and get better deals on the chip. Hearing aids, for example, might be millions of sales, or hundreds of millions eventually.

It's never that the GA 144 was an unusable architecture, but just limited use. It was better than sea forth, which BMW was looking at using.

But, what Rick said, is essentially the truth. The amount of memory per node is small. I can see that this is alright, it's just that it needed a node or more with a full execute memory bus to do things that are suitable for that. I can also say, that
256 words per node, is better for me.
But, it would have been better if versions with more execute memory per node were also used, up till a full 256k per node on chip. Reducing the number of cores cores to fit it (obviously, any
256k word array version, so going be a bigger chip). I also recommended a way to
load in code segments into node memory automatically and like a cache. What it also needed, was an auto way to route on chip communications, reducing the code per node and other things. Unfortunately, things got stuck. Some auto execute serial
embedded memory bus, and modern communications standards would also help USB 3/4 data rate support, ethernet, wifi and USB video HDMI/lcd support would have helped. If somebody does lnt use them, not much is wasted as those pins can be retasked, and
you are only providing physical interface support, which the nodes, and block transfer hardware, run.




--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 7:01:23 AM UTC-4, none albert wrote:

You remember a list of Pintaske, mentioning the exploits of the
GA144 crew? Nothing of interest there. (Not the fault of Pintaske).
I had some hopes that now the patent troll Patriotic is no
longer involved, things would light up, but apparently not.

Then I remembered an interesting talk, I came across the other
day in youtube. So I'd share it with you.
Impossible to find back! No plausible keywords help.
I literally have to wade to all youtube films I viewed over the
last few weeks. Then I found it:
https://www.youtube.com/watch?v=xYWa2C2_7H0
hidden in the non-assuming title "SVFIG 17Nov2018 Afternoon"
that goes on rambling for more than 3 hours.
If you have patience than you find it at 1:09:
A real application! A heart rate monitor!
I concede that heart rate monitors comes free with smart watches
nowadays, but it is really the first time I came across
a description that some one has actually done some with the GA144.

(I discount rumors that Jeff Fox apparently did something with
hearing aids. )

It is interesting that the claim discussed on a 50+ thread
in electronics.design that apparently the GA144 can use
a 32 Khz watch crystal hanging off one of its ports
without any passive or active components needed,
is true.

If you are talking about science.electronics.design, that was most likely myself. I would not have claimed to have made this work, but reported on the work of others. I recall that Chuck tried this with a 10 MHz crystal and failed. This would be based
on a timing loop to get started and the timing variance from PVT (process, voltage, temperature) was too great.

I want to say someone made the 32 kHz crystal work, but I don't recall who. It worked by using a single I/O pin as both input and output. The pin was pinged either high or low, or maybe both alternately, based on the input crossing the threshold of the
I/O pin. The pulse has to be very short to prevent overdriving the crystal and timed correctly to produce a stable oscillation. No one that I'm aware of, ever did a proper analysis of this circuit, rather, they just started playing with the idea.

I had some number of people in s.e.d attempt to apply classical oscillator theory to prove it could not work. That seems a bit silly, but maybe ok if you can figure out how to treat a digital I/O in this context.

--

Rick C.

+ Get 1,000 miles of free Supercharging
+ Tesla referral code - https://ts.la/richard11209

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

In article <5973f2ee-f79c-4741-8810-2265e334870fn@googlegroups.com>,
Hans Bezemer <the.beez.speaks@gmail.com> wrote:

On Monday, September 12, 2022 at 2:20:25 PM UTC+2, Hans Bezemer wrote:

On Monday, September 12, 2022 at 1:01:23 PM UTC+2, none albert wrote:

A real application! A heart rate monitor!

I don't know if you consider this "an application", but I found this:
https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

.. and this one: https://dl.acm.org/doi/10.1145/2666356.2594339

I consider those click bait for hedge fund investors who want to
speculate in patent extorsion, so not really an application.

Hans Bezemer

Groetjes Albert
--
"in our communism country Viet Nam, people are forced to be
alive and in the western country like US, people are free to
die from Covid 19 lol" duc ha
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 8:36:09 PM UTC+2, none albert wrote:

https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

[..]

I consider those click bait for hedge fund investors who want to
speculate in patent extorsion, so not really an application.

Tobias Schneider works at NXP.
Moreover, they wouldn't have printed a table like this:

Platform Time [us] Energy [uJ] -------------------------------'-------------'--------------
GA144 (0.18 um) 37.9 0.90
AVR ATtiny45 [5] 455.7 19.2
MicaZ software [35] 885.8 19.16
MicaZ HW-assisted [35] 350.6 26.82
Many-core proto (65 nm) [18] 0.126 0.198
ASIC (0.13 um) [14] 2140 0.051
Low-power ASIC (0.13 um) [13] 1.23 0.005

I think I would go for the low-power ASIC.

-marcel

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Monday, September 12, 2022 at 3:33:01 PM UTC-4, Marcel Hendrix wrote:

On Monday, September 12, 2022 at 8:36:09 PM UTC+2, none albert wrote:

https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

[..]

I consider those click bait for hedge fund investors who want to
speculate in patent extorsion, so not really an application.

Tobias Schneider works at NXP.
Moreover, they wouldn't have printed a table like this:

Platform Time [us] Energy [uJ] -------------------------------'-------------'--------------
GA144 (0.18 um) 37.9 0.90
AVR ATtiny45 [5] 455.7 19.2
MicaZ software [35] 885.8 19.16
MicaZ HW-assisted [35] 350.6 26.82
Many-core proto (65 nm) [18] 0.126 0.198
ASIC (0.13 um) [14] 2140 0.051
Low-power ASIC (0.13 um) [13] 1.23 0.005

I think I would go for the low-power ASIC.

-marcel

130 nm won't give you many more CPUs than the GA144, maybe twice? I think the GA144 was 180 nm. I think it gets really interesting when you can have 1000 or 10,000 CPUs. It also needs more comms somehow. That's always a problem in multiprocessing,
effective comms.

Any idea what the power level might be per processor on that process? I want to say the GA144 was around 1 mA per CPU running flat out, but it may have been more. Let's see, the chip could reach 1W, IIRC. So around 7 mW per CPU. At 1.8V that's more
like 2 mA. Get that down to 100 uW per CPU flat out, and you've got a true low power chip. 1,000 of those will fit in a 100 mW envelope and mostly run a lot cooler.

How does power scale with feature size? I think 130 nm was a 1.2V process, no? That helps power a lot, 2x! Now find another 35x, lol!

Any idea what is realistic at 130 nm?

--

Rick C. (Lorem Ipsum)

-- Get 1,000 miles of free Supercharging
-- Tesla referral code - https://ts.la/richard11209

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Tuesday, September 13, 2022 at 5:33:01 AM UTC+10, Marcel Hendrix wrote:

On Monday, September 12, 2022 at 8:36:09 PM UTC+2, none albert wrote:

https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

[..]

I consider those click bait for hedge fund investors who want to
speculate in patent extorsion, so not really an application.

Tobias Schneider works at NXP.
Moreover, they wouldn't have printed a table like this:

Platform Time [us] Energy [uJ] -------------------------------'-------------'--------------
GA144 (0.18 um) 37.9 0.90
AVR ATtiny45 [5] 455.7 19.2
MicaZ software [35] 885.8 19.16
MicaZ HW-assisted [35] 350.6 26.82
Many-core proto (65 nm) [18] 0.126 0.198
ASIC (0.13 um) [14] 2140 0.051
Low-power ASIC (0.13 um) [13] 1.23 0.005

I think I would go for the low-power ASIC.

-marcel

I guess that was done early last decade. The energy and performance ranges of the smaller
processes are the sort I'm after, but this gives you no idea of what a 130nm and 65nm GA is
going be like. The selection is confusing too. If only they had the same processor
implemented at 65, 130 and 180nm high and low energy, it might give an idea of how much
these figures benefit from shifts down in fab node size (plus enhancements).

But seriously, they could stack psram dies in package on top of the processor chip, for some
edge cores with full execute bus, for some great potential. Because, the GA uses so little
energy compared to the memory dies, you could run it full speed flipped with the memory dies
on top being air or actively cooled. Pseudo SRAM being cheaper dram with sram interface
and on chip refresh circuits.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 13/09/2022 4:36 am, albert wrote:

In article <5973f2ee-f79c-4741-8810-2265e334870fn@googlegroups.com>,
Hans Bezemer <the.beez.speaks@gmail.com> wrote:

On Monday, September 12, 2022 at 2:20:25 PM UTC+2, Hans Bezemer wrote:

On Monday, September 12, 2022 at 1:01:23 PM UTC+2, none albert wrote:

A real application! A heart rate monitor!

I don't know if you consider this "an application", but I found this:
https://www.cs.bham.ac.uk/~oswalddf/publications/ga144_journal.pdf

.. and this one: https://dl.acm.org/doi/10.1145/2666356.2594339

I consider those click bait for hedge fund investors who want to
speculate in patent extorsion, so not really an application.

If extortion be the goal what's better than a pharmaceutical company
where the client (typically govts) have no choice.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)