Welcome to the home page of General Application Processing™. Read about the creation of a Flexible System Architecture™, a hierarchically parallel processing system providing very high security and fuel efficient computation.




Screen shot of the FSA™ Simulator thirteen cycles into a CORDIC computation of sine & cosine. Input: 45 degrees. Output: cosine is at top of P1 at level C3, sine is just below it. Both values are .70703125 in a fixed point format with thirteen bits to the right of the binary point.





THE DIGITAL WORLD AS OF 2020

“Surveillance capitalism” - that's the snarky description by Shoshana Zuboff of our current Internet. Various people and organizations are trying to change this mess, including Tim Berners-Lee, the inventor of the Internet.

My question is: Do we need new hardware on which to run this vast network?

I've created the Flexible System Architecture, a different approach to digital computing.

The FSA's internal instruction size is 16 bits. Why so small? Mainly, to save energy. The carbon footprint of the Internet is growing quickly and becoming disturbingly large. Sooner or later this will need to be counteracted. Digitally switching 64 transistors simply takes more power than switching 16 of them. The same goes for rows of any kind of logic circuit.

Energy use is one problem with the current Internet. The second is lack of security. The FSA is an architecture with security built into both its instructions, and the hardware layout implementing it. We are heading into proprietary issues here, But I will say the FSA consists of small sequencers with a minimal circuit footprint. They are meant to be interconnected in a pyramidal parallel system, which allows for higher levels to be protected by the actual physical wiring.

There is a third problem with the current Internet: A lack of idealism. The original Internet was created with a dream of assembling a World Wide Web of intellectual collaboration. Now it seems mainly to exist to sell us something.

Going further back, the original personal computers were created with a dream of liberating users from the tyranny of sitting in offices, typing on dumb terminals connected to mainframes. To a degree, the financial investors in PCs shared that dream. Now investors seem to only be concerned with quarterly reports and growing startups too quickly for their own good. Pleasing Wall Street and cashing out are the new standards. "Greed is good," as Michael Douglas pronounced in the 1987 movie, "Wall Street".


THE FSA RIGHT NOW

The Flexible System Architecture is a work in process. The most important subset of its machine instructions - what I call "Active Instructions" - are well simulated and operate as designed. They make use of the pyramidal security layout to keep higher levels safe from state changing by lower levels.

Another "Passive Instruction" subset runs on a separate bus, sending messaging bit groups throughout the architecture. These aren't yet as fully simulated, but I believe they are well fitted to implementing neural networks. A third bus is for data, which need not be limited to 16 bits.

The aforementioned Simulator [See the graphic above] is bare-boned, yet has served me as an effective development system for many years.

Work has also been started to develop a lightweight human readable language to be layered on top of the machine instructions.


POSSIBLE APPLICATIONS

I often describe the FSA as a "Control Architecture." I began work on it by way of a fascination with the idea of controlling billions of transistors on a chip. The beauty of that is if it can control those, it can control anything else. This makes it suitable for any such application, which these days particularly means the Internet of Things.

I must admit I'm not filled with deep warm fuzzies toward IoT. I recently ran across a headline about "smart umbrellas." I didn't click on the link - I was too busy thinking, "Seriously?" The hype has spurred any number of frivolous undertakings, but the worst result is that devices and drivers are being rushed into development while security is largely a second thought.

My personal lack of enthusiasm regardless, IoT would appear to be a natural fit for the FSA. The market is definitely there. Depending on how many individual blocks are stacked in a pyramid, the architecture could handle everything from simple control to complex edge processing, all while saving power and repelling hackers. Maybe ultimately the FSA's most important role could be to save IoT from itself?

While IoT could well be the ultimate "killer app" for the FSA, it is certainly suitable for other employment as well. After all, the URL of this very website is a triple portmanteau of 'General Application Processing'. The basic architectural unit is a full function computer in its own right. Of course there are many processing tasks where 16 bits alone don't cut it. But the FSA does not have to be the only type of processor in a system.

Each standard FSA unit has its own internal hierarchy, and various kinds of IP can be embedded within it, including commercial processors and coprocessors. So perhaps a better description of the FSA is to call it a "Coordination Architecture," one able to efficiently oversee tens of billions of transistors at the board level, expanding beyond the chip level mentioned above.

If the firmware of a 'foreign' processor in a lower building block within the hierarchal pyramid ends up corrupted by a virus, it would nevertheless be guarded by the web of the secure FSA instruction set it resides in. The hacked subsystem can easily be isolated, and possibly even be disinfected by algorithms running in higher levels.

Another moniker I've given this architecture is "Decision Engine." Besides a standard ALU, it contains instructions and data paths for non-standard (even user-configurable) logic operations for speedy reaction to changes within the larger system [See the GIF representing a single standard FSA unit below].


WHITHER THE FSA?

A lot has been covered here in a short time. I believe I've presented a unique system design. It it marketable? Hard to say.

I'm an inventor, so naturally I want my great idea to take over the world. But does the Internet really need a low level, tiny-message passing system as its new backbone? I would say the FSA aims for such fine tolerance that moving beyond board level control is unlikely. But maybe. We shall see.

One other dream common to the creators of the original PCs was the vision of them ultimately becoming small enough to be truly personal. That indeed happened (although whether anybody back then foresaw them also being phones is doubtful).

These days, we appear to be returning to the bad old days of the mainframe, only now it's called The Cloud. A collection of wannabe monopolies are competing for dominance, hoping we'll store our digital lives there. I consider this a bad trend. If I personally have a dream for the Flexible System Architecture it would be to ultimately implement a new class of laptops, tablets, and phones.

Security would be the main selling point of such FSA based devices. ID numbers, passwords, medical information and the like would be tucked away behind the top of the pyramid, while outside communication is plugged in at the bottom.

Yet I would like it to do more. There should be a simple hypervisor that gives the user true autonomy: the ability to really turn off apps, tracking, whatever. One shouldn't have to "root" a device to gain administrative control over it. Private should mean private.

In general, I'd like to see a better balance between the processing available in the cloud versus personal devices. Obviously, huge farms in the cloud are going to have more computing power than any portable unit, but I see no reason that the unique features of the FSA should not be put to the test to address this difference.

I described some of these features above: non-standard and user-configurable logic operations, a class of instructions fitted to implementing neural networks, and the fact that each standard FSA unit (i.e., block in a pyramid) has its own internal hierarchy for embedding various kinds of IP - which could include neuromorphic circuits.

Neuromorphic hardware is made up of individual "neurons" composed of few transistors, which means that high numbers of them can be packed together on silicon to create a neural net. The event-driven processing of neuromorphic systems is extremely efficient compared to conventional Von Neumann sequencing, And of course, efficiency is one of the aims of the FSA.

Imagine: a mixture of pinpoint neural networks operating within a software based neural system using FSA instructions - call it "AI For the Rest of Us."


To me, the digital future comes down a question of self-determination. Is our personal identity going to be subsumed into a larger commercial authority, or do we hold on to our individuality?

The choice is ours. It is always ours.





            Orignal static taxonomy graphic from EDN Magazine *







About Previous
Home Page
Original
Home Page
Security
Thoughts
Table of Contents
Ongoing Writing   —— Essays ——   "The Perfect Language"


Notes on the links:
About — Including a link to my resume.

Previous — From 2016, a different view of the FSA, still worth reading.

Original — "The Chip for the New Millenium" ... A bit premature.

Security — My early thoughts about a hierarchically secure system.

TOC — There's a more complete Sitemap available, but this is more up to date.

Essays — My musings on math, science, and computer languages. Some of the language blogs were speculative, and have turned out to be wrong. Two articles worth highlighting:
An overview of Binary Decision Programming , a subset of the FSA instructions.

An explanation of the CORDIC computing system illustrated in the graphic at the top of this page.





Bob Loy, Founder


* From EDN magazine, Issue 11, 6/11/2009 — "Tee up your multiprocessing options" by Robert Cravotta