GENERAL APPLICATION PROCESSING
FLEXIBLE SYSTEM ARCHITECTURE
|Executive Summary 2010 Ongoing Update|
I got fairly early into web design back when the Internet was still a friendly, collegial place to visit. I shared the basic idealism of the time; very few people tried to hack one another, and gentlemen certainly did not read each other's mail. The World Wide Web was going to create a better, wider world. What a difference a few years of "internet time" has made! Earth is a better place thanks to it, but WWW could now stand for Wild West Web, filled with rustlers and claim jumpers.
I now find myself in the almost schizoprenic position of having created a computer architecture whose most important market just might be computer security. Although I originally wrote a page about security applications as an addendum to this executive summary, the reader may want to go there now and then come back to finish this page.
Differentiation is one of the names of the business game. Does the FSA enter new computing territory? Just above is an original taxonomy graphic from "Tee up your multiprocessing options", an article from the June 11, 2009 issue of EDN magazine.
Beside it is a modified graphic showing the "sweet spot" of the FSA. Actually, this architecture can easily cover the whole left third of the graphic, but its sweet spot is decisions, decisions. Is there a market for such? Usually any performance increase in computing hardware makes somebody somewhere happy enough to pay for it. Still, some market research is definitely due.
Nature of the Product
The FSA is a very adaptable, modular computer design (or "architecture") able to be fabricated as various types of microprocessing components and computing system elements.
One application it lends itself to is the implementation of highly secure home (and mobile) computing. See the text link in the introduction above, or read to the bottom, then click the middle link.
The Flexible System Architecture can operate as: a stand-alone computer, a parallel processing element, or as a set of distributed control cores within System-on-Chip (SOC) and other ASIC development. This flexibility can be applied to tasks ranging from simple appliance control to supercomputing. Between those two extremes are toys, office machines, wireless communication, and automobile control systems, to name a few.
www.electronics.ca mentions emerging "end markets in industrial lighting, alternative energy, and IC cards" and "the importance of alternative energy, including wind and solar" which has "pushed consumption of microcontrollers for advanced motor control and energy monitoring."
Here in 2010, maybe the most exciting new frontier for digital electronics is that of web appliances. The coming revolution in cloud computing may not make the desktop computer obsolete, but in overall numbers of devices connecting to the Internet, the share owned by traditional PCs will dwindle to a small minority.
Dozens of specialized devices, phones, PDAs, Kindles, what have you, are being and will be created that bypass the need for either a monopoly operating system or a monopoly processor architecture. "Wintel" will be replaced with a plethora of new systems and applications. In this rapidly changing environment these devices will need to be quickly developed and easily reprogrammed. There's no reason the FSA can't be a player here. One dirty little secret is that many of these products have potential security holes gaping wider than those of standard desktop systems another job for the FSA.
Here is perhaps the hottest potential niche for the FSA:
Silicon computing is still doubling in speed every two years or so, but nevertheless can't keep up with the amazing bit rate (forget giga, think tera) increases in the fiber optic networks linking our world together. Yet that data still needs to be formatted, compressed, encrypted and/or switched to get where it's going.
Because of this mismatch between "control plane" and "data plane," most of the control circuitry is still being expensively custom designed. For a very good overview of this situation, see: Software: the Achilles' heel of network processors?, an article from the April 2002 issue of EDN.
Another timely product to create using this architecture might be an ultra-fast Java engine. Hitching a ride on Java would also greatly reduce the burden of creating system software and applications for it.
When a product in the above categories needs to be even smarter or faster, units can be ganged into parallel processing systems, or specialized circuits can be added to the basic design. These options also open up more complex application possibilities.
While it is to be hoped the FSA will find market share in areas now occupied by traditional computer designs, higher profit will be achieved when it breaks new ground due to its unique features. There's a need for a non-custom solution with rapid design cycles, fast processing speed, inherent parallelism, and a small silicon (or GaAs? Graphene?) footprint.
Sales and Company Growth
There are several sales areas which can be addressed by this technology or its offshoots. A quick list:
Create and market a standard microprocessor, initially farming out actual fabrication of it to a silicon foundry
Sell or lease development systems associated with the micro family
Write programs to run on the micros in end-users' products
Design custom circuits to add intelligence to end-users' products
Actually manufacture the microprocessor family
Develop and manufacture consumer/industrial end-products
No complex, programmable integrated circuit gets sold without an accompanying development system. In order to make "smarter" end products like those described above, the micro that's contained in them must run some sort of program. The development system makes it easy to write such programs; in fact, they would be practically impossible to write otherwise.
Interestingly enough, this same system must be developed as an engineering tool in order to create the microprocessor in the first place, so two very different marketable items are produced at the same time by the original startup.
Many OEMs (original equipment manufacturers) lack the personnel or expertise to use a development system to write the programs they need, so they hire outside help. This creates an opportunity: The same engineers who developed the architecture and its tools can go to work writing the programs and/or designing custom logic for customers.
Thus, the startup company or division evolves from a design center into a high-tech, high-ticket service organization. This can create a cash flow which will help the bottom line while waiting for microprocessor sales to grow, and will allow for further refinement of both the micro family and the development system to increase their competitive advantages.
When sales become large enough, the original company can build a fab line and keep all the profit from its microprocessor and ASIC sales. Finally, the company can conceivably spin off a subsidiary to develop unique end products in the consumer and industrial sectors.
Thus, marketing a micro family is not limited to the chips themselves. Over the short range, a minimum package ready to sell would include the following:
An integrated circuit built around the basic architecture
A software development system consisting of:An easily ported simulator of the IC which featuresResponsive technical support for all of the aboveA graphics displayA symbolic assembler
A user-friendly monitor
A symbolic debugger
High-class documentation for all of the above
Size of Market; Growth of Market Potential
The size of the 16 bit microcontroller (MCU) market seems to have leveled off over the past decade. It almost reached $14 Billion in 2007, but in the current world economy it has shrunk back to the $12 Billion range it inhabited in the late 1990's.
Some caveats about applying this data to the Flexible System Architecture should be mentioned: The FSA not aimed at being only (or even primarily) a microcontroller. It is more powerful than the average 16 bit micro, and its main target application is to be a core processor within a bigger system on a silicon chip. Still, one has to look somewhere to at least approximate the size of a potential market, and the microcontroller market is the easiest about which one can find publicly available information.
A second caveat is that the data from a decade ago came from the World Market Share Reporter, while the current data comes from the www.electronics.ca website. It's possible that their metrics differ somewhat.
With that in mind, we find Motorola's spinoff, Freescale Semiconductor, has seen its share drop from 17% to 11%, but it still holds on to second place. The new leader at 20% is Renesas Technology, which was formed from a merger of the Hitachi and Mitsubishi semiconductor divisions, both of which were previously ranked in the top ten.
The important fact is the continuing fragmentation within the 16 bit market. In fact, the top ten companies now leave a 23.5% share unclaimed, as opposed to 17.1% ten years ago.
Being a "do-anything" design, the Flexible System Architecture can be applied across the whole spectrum of digital technology; its potential market growth IS the market growth of computer technology in general.
Even a 1% penetration of this volatile, easier to enter market would yield a $120 million company or division, supporting nearly 600 employees.
IBM reports an estimate predicting the ASIC market will reach nearly $40 billion by the year 2001 (source: www.chips.ibm.com).
In every single area mentioned, the FSA can have a direct impact.
*source: "World Market Share Reporter 1997 - 98," values
rounded up to the nearest billion
**source: World Semiconductor Trade Statistics organization of the Semiconductor Industry Association
The competition in these well established markets is truly fierce, definitely making for an uphill battle.
The 16 bit realm in general is getting squeezed from both directions. 8 bit processors are getting added functionality while 32 bit processors are getting smaller and cheaper, sometimes with a stripped down 16 bit version also available (as is the case with various ARM offerings).
Some companies, like Freescale and Microchip Technology, have product lines covering all three word sizes. A 16-bit-only offering would have to have something special to compete against such lines.
Engineers have been making cores for a long time. More importantly, they've been developing software tools to help designers create with those cores and easily insert them into new products. The industry has a huge head start over the FSA in the development of such tools. Catching up in this area would probably take the biggest share of investment capital in the long run.
The saving grace for the FSA is that these markets, although well-established, are not static. Digital electronics continues to break new ground. There seems to be no end to what is almost exponential growth in new applications and products.
Selected Specific Competition
The new top ten in 16 bit MCUs:
More specific to the FSA's targeted use for internal core processing, the following companies use the Embedded Microprocessor Benchmark Consortium's CoreMark Core Benchmark test and rating system:
Microcontrollers: These companies have a market share of 5% or more: Motorola, NEC, Hitachi, Texas Instruments, Intel, Mitsubishi, Lucent, and Philips
Application Specific Integrated Circuits: LSI Logic, IBM, Motorola, NEC, Lucent, VLSI Technology, Integrated Device Technology
Network Computers (the Java connection): Oracle, Tektronix (the NC200 series)
While a powerful stand-alone microcomputer system could be built around the Flexible System Architecture, its preferred application is as an embedded core within a larger System-on-Chip environment. Indeed, in a very large system, FSA cores could be distributed in a hierarchical fashion as needed, a scheme simply laid out because individual FSA sequencers can easily find each other within a larger process address space supported by the instruction set.
Even in a system with other kinds of embedded cores, the FSA would be the best choice for the top-level control core. There are several reasons for this. The basic FSA standard sequencer is minimally pipelined, meaning it can respond rapidly to tests and context switches. Its response to test bits in particular is extremely fast. It is as easy to program state machines, for example, in FSA native code as it is in any 8 bit processor.
The FSA has an enhanced ALU with special logic functions over and above what is standard in the industry. This can provide for greater code density, letting a given program fit in less memory space, which in turn allows it to complete its function in fewer instruction cycles.
The FSA instruction set designed from the ground up to control digital circuitry. In addition to the inter-sequencer communication and fast test response already mentioned, there is a proprietary mechanism for propagating control signals that helps to avoid data bus contention, allowing the circuits being controlled to talk to each other with minimal interruption.
Finally, while the ideas expressed there do not necessarily apply only to the Flexible System Architecture, the FSA does lend itself particularly well to the general system outlined in secure.htm.
Proprietary Advantages: In April 1998, a Provisional Patent Application covering the full architecture and specific internal circuits was filed with the U.S. PTO. This protection was reinstated in August 1999 for another year. In addition, the instruction set should be copyrightable.
Competitive Advantages: Here is a quick list of the general advantages the FSA can provide to those looking to create the next generations of intelligent products:
Faster product design cycles
Lower system development costs
Lower system manufacturing costs
Lower cost of final products
Final products run faster
Final products use less power
Possible to provide more options to the system designer
Easier to modify and upgrade products
Easier to repair products
Products can be extremely miniaturized
Possible to provide more options to the end user
Architecture should be superior in attacking currently intractable computing problems
In most cases, a company looking to travel from initial idea to final product will be in a race with their competition. Therefore, the most important item on the above list is probably "Faster design cycles." This architecture and its attendant development system will have several inherent advantages in turning a product from idea to market more cheaply and efficiently than competing alternatives, yet allowing creation of an innovative and superior end-product.