The patent system is to innovation what politcal correctness is to free thinking - death! To completely abolish the US patent system would be to give the real inventors in this country a chance to compete, long term, with the rest if the world.
I actually·just emailed EE Times earlier this week to·suggest they give my free subscription to some imminently qualified IP attorney, instead, since they seem to think these people·are the future. I·imagine (or like to imagine) that they got a lot of backlash from readers who complained about their March 20 issue where they glorified patent trolls as a 'New breed of innovator'. And, in their Crosstalk section where people write in, they published two letters, obviously from attorneys, who said that the US patent system is just fine after 200 years, and it shouldn't be changed. Right. This article you pointed out from April 3 was probably some compensatory reaction.·They've been leaning further and further left, anyway.·They emailed me back and said·"We are sorry you are so displeased with our publication." I am, too. I used to love it and recommend it to everybody. Here's the article, by the way. It might be good to be sitting down,·perhaps on the toilet:
Phil Pilgrim (PhiPi) said...
As a followup to an earlier discussion in this thread about the patent mess in this country, here's a link to a recent editorial in EETimes ("No Patents. No Kidding" by Rick Merritt):
I believe Mr. Merritt's argument has merit. Our current system is broken; it rewards the wrong people; it stifles true innovation, rather than rewarding it. Why not just get rid of it entirely?
-Phil
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/21/2006 5:58:10 AM GMT
Is there any reason conceptually that you couldn't create a Propeller chip that had hundreds, or thousands of cogs? Or even islands of cogs that had a local hub, with all islands being managed by a central hub?
If it's doable, it would be pretty neat to develop a microkernel operating system where each process was run on an individual cog, or handled by islands of cogs.
Kevin Wood said...
Is there any reason conceptually that you couldn't create a Propeller chip that had hundreds, or thousands of cogs? Or even islands of cogs that had a local hub, with all islands being managed by a central hub?
If it's doable, it would be pretty neat to develop a microkernel operating system where each process was run on an individual cog, or handled by islands of cogs.
That's possible, but it would take a very advanced process to be able to achieve that level of integration. I've heard about chips that have several hundred cores, but those cores·are probably very simple relative to·general-purpose processors. But, yeah, you could make all kinds of tree structures with any type of heirarchy you want. I've heard people say that nVidia chips are almost 'alien' technology, being able to execute 1,000 line programs loaded with floating-point math for every pixel on a high-res screen at full refresh rates. Ponder that. It overwhelms me, but it must break down to simple things. At the bottom level, everything is simple.
That thing sounds like a real monster! This must have been in the mid 1980's, or so? I bet that thing took a lot of power, too. Things have gotten smaller and faster since then, eh?
I'm glad you are into this parallel processing stuff, because I find it intriguing, too. If you want a 'sponsor' to try this out we could volunteer some chips. We don't have $40 million, though, or·maybe $100 million in today's dollars.
pjv said...
Hi Chip;
In the mid nineties a company I eventually owned had developed a multiprocessor super computer. In fact, for a brief period we could boast about having the word's largest supercomputer with 1,036 Motorola 6800 processors; 4 per each of 256 boards 18 inches square (plus some housekeeping processors) in a total of 18 cabinets, each about 30" x 30" by 4 ft high. It filled a whole lab, and cost investors over $40 million in development work. The 1 Gigabyte "hard drive" was a cabinet full of 1 bit streaming drives, 8 bits wide plus a parity drive. Each "bit" drive cost $10,000, so the cabinet was about $100 K.
The target applications were compute-intensive; 3D seismic computing, weather and pharmaceutical modeling etc. As I recall for some demonstrations, it did a wonderful job on Mandelbrodt displays.
The toughest part was the software to keep it all coherent... the processors were not hardware lock-stepped, it was all achieved by software. As I recall there were a number of patents involved.
As we were always playing "technology catch-up", we went on to build a single cabinet of 64 Motorola 68000 processors, but these machine well ahead of their time from an application perspective, and we could not find many buyers at $5 million a pop, so eventually the company became defunct. I still have a bunch of these in a warehouse; for what I don't know!
As you can see, parallel computing is one of my favourite subjects, and I applaud you for suggesting the subject with the Propeller in mind. It could be a lot of fun..... but also a lot of pitfalls if anyone expects to "go commercial" with them.
Can you put interprocessor comms (and power, and, and, and...) in the scribe lines on the wafer, and reanimate the old wafer-full-of-chips array plan? Wafers thet get diced lose the IPC, and you can keep an occasional wafer to play massively parallel games with [noparse]:)[/noparse]
Chip, I work for one of the 3 major Litho tool suppliers (can you guess which one if I am in Europe?), I have the same question as steveW, out of curiousity. Can you divulge suppliers? I have been wondering if our tools have had a part in this (and can almost guarantee that we did[noparse]:)[/noparse] )
The problem is yield. Only ~84% of the dice are good. The remaining ~16% have some defect which renders them not fully functional. So, on a wafer, we don't know which dice are good until they are singulated and tested (we test only packaged parts). An area of a wafer containing multiple dice is bound to have some defective ones.
The die size is 53 square millimeters - HUGE. If we hadn't used a full-custom approach, this would not have even been practical·in a·0.35um process.
SteveW said...
Can you put interprocessor comms (and power, and, and, and...) in the scribe lines on the wafer, and reanimate the old wafer-full-of-chips array plan? Wafers thet get diced lose the IPC, and you can keep an occasional wafer to play massively parallel games with [noparse]:)[/noparse]
(Is the die size a secret, by the way?)
Steve
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/21/2006 5:25:45 PM GMT
Chip Gracey said... ·I've heard people say that nVidia chips are almost 'alien' technology, being able to execute 1,000 line programs loaded with floating-point math for every pixel on a high-res screen at full refresh rates. Ponder that. It overwhelms me, but it must break down to simple things. At the bottom level, everything is simple.
Here's an interesting application of that "alien" technology. Database sort performance increased 10x by·using a GPU instead of a CPU. Makes low-end CPU+GPU systems comparable to high-end CPU-only systems on sorting tasks:
We are using austriamicrosystems' .35um double-poly CMOS process. They were the only foundry we encountered·that·was eager to work with newbies like Parallax. It has worked out really well, though. Their quality has been excellent and it's been a pleasure to work with them. They are at least as much human as they are machine.
parsko said...
Chip, I work for one of the 3 major Litho tool suppliers (can you guess which one if I am in Europe?), I have the same question as steveW, out of curiousity. Can you divulge suppliers? I have been wondering if our tools have had a part in this (and can almost guarantee that we did[noparse]:)[/noparse] )
Chip,
By having X propellers enumerated each time they power up, will they enumerate to the same position each time? Or do you know that answer.
As for patents, I concur with both your views and feelings. But, I also believe that trade secrets have more discrete value than patents anyway. Last November the USPTO granted a patent to a man who had be filing since 1988. I was for the thought process to pay a manager. No formula, nothing other than a thought. It was reported in a UK programming journal in January. The patent system in the US has reached a point where it has so little value, it's not worth talking about. What lawyers haven't destroyed politics has. But I think it's going to get worse before it gets better. Rolling blackouts in Texas this week and people in your fair state are putting their jewelry in hock to buy gas. Hardly signs of a progressive society. But the real question here is, how expensive will gasoline get before the politicians start admitting how bad things are?
You are mentioning 2nd generation Propeller chips. Are they already on the drawing board? Also, what is the fastest clock input you have had the chip working with. I havn't seen the chip but it appears that it would handle more speed.
Congrats on you new product. I sent you an email weeks ago but you were probably too busy to see it. I'm looking forward to writing about it.
Oliver H. Bailey said...
Chip,
By having X propellers enumerated each time they power up, will they enumerate to the same position each time? Or do you know that answer.
Yes, you bet! Otherwise, it wouldn't be worth much. Their enumerated IDs can tell them where they are within an array, or maybe what their I/O's are connected to. Repeatability is paramount.
As for patents, I concur with both your views and feelings. But, I also believe that trade secrets have more discrete value than patents anyway. Last November the USPTO granted a patent to a man who had be filing since 1988. I was for the thought process to pay a manager. No formula, nothing other than a thought. It was reported in a UK programming journal in January. The patent system in the US has reached a point where it has so little value, it's not worth talking about.
I think trade secrets are more valuable than patents, too. About ridiculous patents, one guy recently got one for an anti-gravity spaceship. I'd like to see him build one of these. I heard of some lawyers who had a contest to see who could get the dumbest patent. One guy got a patent for the comb-over (when you're bald and you grow your side hair long and then comb it over the top of your head), and another guy got one for swinging sideways (by moving your feet left-right-left-right on a swing). I want·to get one for 'the thought process to get a patent' so that I could put an end to this whole thing.
What lawyers haven't destroyed politics has. But I think it's going to get worse before it gets better. Rolling blackouts in Texas this week and people in your fair state are putting their jewelry in hock to buy gas. Hardly signs of a progressive society. But the real question here is, how expensive will gasoline get before the politicians start admitting how bad things are?
This subject is·like Pandora's box.
You are mentioning 2nd generation Propeller chips. Are they already on the drawing board? Also, what is the fastest clock input you have had the chip working with. I havn't seen the chip but it appears that it would handle more speed.
The next generation propeller is being laid out, although there is some serializer design to do. The current chip runs at up to 112MHz, but 100MHz is safer for room-temp apps. It will be officially·rated for 80MHz, though.
Congrats on you new product. I sent you an email weeks ago but you were probably too busy to see it. I'm looking forward to writing about it.
I'm sorry about that. I don't remember seeing it.
Oliver
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/22/2006 8:53:38 AM GMT
crumbs! That's a lot of silicon to be generating so little heat. Nice going [noparse]:)[/noparse]
>The problem is yield. Only ~84% of the dice are good
That's always been part of the fun of waferscale stuff, though - skipping the bad'uns, but still having the array be useful. (for some value of 'useful'...)
Still pondering what my killer app for Propeller is going to be. Too many choices - curse you and your excessively capable silicon [noparse]:)[/noparse]
Still pondering what my killer app for Propeller is going to be. Too many choices - curse you and your excessively capable silicon [noparse]:)[/noparse]
Steve
You're welcome! Does this have anything to do with Crypto-Sporidium?
Trade secrets or just plain copyright with an encryption scheme included may be useful to Parallax.
I delved back into reading about FRAMS microcontroller support chip which includes a 64 bit 'serial number' feature. It seems that it might be used to decode an encrypted software scheme.
Thus, you might protect future versions of your IDE from being used with other people's copies of your work.
I know it is a lot of might and maybe. I just don't have the know-how to achieve it myself.
Sadly, I suppose it would mean having a registered chip owner with a registered IDE.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"When all think alike, no one is thinking very much.' - Walter Lippmann (1889-1974)
······································································ Warm regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
I can't imagine anyone copying the Propeller or Spin directly.
The Spin interpreter is hardcoded for the Propeller machinecode, and any attempt of adapting it on another processor will demand a complete rewrite.
Besides, the cloners who have copied the Basic Stamps have a lot of catching up to do if they want to replicate the Propeller, and THAT's the clue, without a multicore chip(like the Propeller or the Cell -- which is supposedly used in the mythical PS3 -- processors) there's no reason to even bother looking at Spin.
Lets face it, the big ones, like Intel and Amd won't be interested in something like the Propeller(Intel, at least, have their own microcontrollers, running with high enough clocks to get the same MIPS out of them to rival the Propeller, and they've always thought that speed is king) and the smaller ones may not want to invest that much into a new architecture before they know it's a viable market, and even if they decided it is, they'd still have to develop their own chip, and as Parallax has proved, that takes time.
Even if they could assign a large group of people, I doubt they could do it in much less than two years, unless they want to end up with a very buggy and unstable product...
and by that time, we'll be ranting and playing with the next generation of Propellers...
The problem with serial-number ICs and encrypted software is that the Propeller doesn't have a protected ROM, so any decryption-software will be available for anyone to copy out and reverse.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Don't visit my new website...
Propeller supercomputing arrays... Lets think about the pros and cons, shall we?
A Few Pros
* With arrays of propeller chips, you could create neural networks.
* Propeller chips do not have operating systems - that is a big advantage over people running code on a pc with a windows xp home os and a P4 2.8Ghz processor (me.)
A Few Cons
* As someone I forget has said: multi-processor systems sum up to less than the·sum of the individual parts alone, because of time·spent·communicating, verifying, etc.
* These MCUs are still, (even in bunches) MCUs. There are other limits (like, is there hardware 32x32 multiplicaton or division?)
Conclusion
I don't think that multi-propeller boards will be seriously implemented, except for a few esoteric projects. It's still fun to think about, though.
IMHO:
Its simple enough for someone to dream of the need for lots and lots of pins.·
Also, it’s simple enough to use clever programming and hardware to provide other "nice" features between more then one controller.
The bottom line is how much effort does the designer want to go through to make a project happen?
For example:· I have no qualms about dedicating a propeller, and some hardware to control some servos, making this a subsystem within my project.
IN essence, this is the same as making an array.· Only, I'm specializing the functions of the subsystems, or atoms.·
Very similar to the human brain; the parts of the human brain that control speech cannot do anything with vision.·
Multi processing is something our brains to all the time, but with specialized interconnections and functions.·
Making an array of propeller chips to perform a give task, a task that leads towards the needs of multi processing, could lead to better performance.·
I agree with most of what was posted here, BUT:· I also believe in specialized functionality.·
Generic routines for each and all propellers, but, each would have it's own reason (code/hardware/object) for being included in the array.·
Just imagine something with the need to control 1152 servos.·
Forget about the power requirements, and other "hard facts", just think about how you would control them.·
In essence, you would load the exact same code into each propeller that would run the servo controlling object,
but EACH propeller would have something unique to it's self, so the object is actually running ALL the servos knows which propeller to send a servo request to.· Array programming coupled WITH specialized functionality.· Knowing redundancy is something of a necessity in any closed system, there would be some duplication of specialized functions, but that's duplication, not replication.·
No, I don’t believe in the statement that all nodes can perform any and all other nodes functions.· It doesn’t provide for specialized functions.·
If something needs 1024 bits of floating-point math, make an object (either hardware and or software), and give it to it.·
But don’t expand your expenses to give this to all other controllers in the array.· This is more then just an environmental or ecnomical issue.
Just because we want every house phone in the USA to be able to talk to every other house phone in the USA, doesn’t mean everyone’s phone is connected to everyone else’s phone, and it doesn’t mean that everyone has to have the same phone. ·
Yes, the idea of Multi Propeller applications is exciting, and would be a worthwhile learning experience for all who take that path, but the truth of it is this:
Given that all propellers from the same·family are created equally, there can be no gains, unless specialized functions are used within each.
....
Thank you for reading.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
1152 servos... hmm... maybe a giant robotic arm that assembles planes and space shuttles. My main point was, even if there is nothing I could pull off the top of my head then, there is something enticing about so much processing power, in such a small package. Although, now that I think about it, computing pi, driving a space rover, and a lot of other junk comes to mind.
The most fascinating aspect of a Propeller array to me is it's ability to evolve over time.
Since individual cogs can dynamically pull in program instructions they could reprogram each other or respond to stimulus.
Imagine a massively multiplayer online role playing game where the nonplayer characters were represented by individual cogs.
Given enough memory space they could remember who they met and whether they were treated well or badly.
You could send them on quests which would be small subprograms that they could load and execute.
If it was done well it would make a nice Turing Test...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
I wonder if this wire is hot...
As I said before, about the space rovers and NNs (neural networks)... and self modifying is another way to achieve semi intelligence... maybe we should start a project for the code to drive a rover. That would be cool [noparse]:)[/noparse]
You know, there are robots which could legitimately carry a propeller cluster.
I'm currently doing undergraduate research in AI at the University of Oklahoma. I actually have built physical platforms (they move but need electronics added for brains) for a trio of giant robots (today, 3, tomorrow, an army) for which a "mobile supercomputer" of an 8x8 or even an 8x8x8 cube of Propellers would actually be appropriate and feasible. The four motors together peak at 100 amps current drain from large lead-acid batteries. The platforms are built with only one design goal: to be able to take more punishment than anyone could possibly expect a robot to survive, and keep coming back - I drove the prototype through waist-high weeds, knocked small trees over, ran it over cliffs, etc. Each robot is strong enough to carry a grown man on it's back; I could add 100 pounds more batteries to them without harming the physical frame or bogging down the motors - it definitely has the room and battery power for 64 propellers. One thing you could do with 512 processors is learning using a Genetic Algorithm. There are also robot control architectures (subsumption, vector summing, etc.) which fit the propeller-cluster hardware better than they fit the singule processor computers they are usually implemented on. However, what I would use a propeller cluster for is machine vision.
I'm currently only an undergraduate, but I plan to use these robots in my future Ph.d. work. I have been funding the project with my own money, so right now I have to settle for 1 propeller in each robot. If I got outside funding for it, and if parallax made such a propeller cluster board, then I would probably put them in these robots. But that's a lot of if's.
His last design was called the 25X. It had 5x5 of his forth engines on it and he claimed pretty much unlimited MIPS.
Seems like the propeller was a smaller realization of this theoretical chip. I will say that spin is eminently more readable compared to forth.
Is the spin interpreter a threaded interpreted language? It would make sense to me, as a forth interpreter is very small. Where each command is made up of a series of commands until they get to primitives which would be snippets of machine code.
I don't know if a multiprocessor board would be useful (right now). The high speed design with so much fan out would be expensive to design. You know how much of a problem the delays were with 8 on a chip die. Just imagine the problems taking it to FR4. I would wait until you see designs with two props appearing in the hobby world before going down this route.
Chip:
Can I ask one question? Are atmel and microchip afraid of you yet? If they aren't then they should be.
Post Edited (LucidGuppy) : 7/27/2006 12:36:03 AM GMT
Does anyone have any links on machine learning, neural networks, or genetic algorithms. It would be interesting to develope one of these with the upcoming hydra for a game.
Hi to all! Just a few points to comment on: 1) A 16 dimensional hypercube is called a Tesseract: The image of wich·is a simple three-dimensional representation reduced one further dimension by being viewed on a flat screen. 2) With regards to the max density that you could logically make A Prop matrix either as individual ICs or as one Die is the same! It dos not even matter what material it is made out of. For example: Even if it were made of tissue paper the results will still be the same. ·Read Scientific American, (can't remember the date, 2004/2005), but it has been calculated and found to the amazement·of the scientific community, (If for no other reason than no one ever bothered to think about it.), that the absolute highest memories density that can be achieved is the same density that is just short of a Black hole. If the density becomes any greater then it becomes a black hole and there for all stored memories are inaccessible do to the extreme high gravity. 3) With the coast of prop chips now being lower: It is possible·to build a 4x4 matrix for $192.00 ···· As for the interface of the chips: Kiss them just I2c them together on P30&31 when p29 goes high then two things happen in parallel. a) a simple I2c/(DS1wire ID&/ switch that turns on the next Prop, or any prop in any order, this is done by sending out a code to that switch in an order determined by·the master Prop, (ID#01, it's the prop·with an eeprom connected on P30&31). So! P29 going high says to the master prop I’m done and ready to be loaded with my new activities. b) When the Master prop sees that P29 on the current prop IDxx is High it then sends it·data to that chip. the other chips wont be looking @ that time and the next or all unloaded chip/s will just be sitting Idle till there individual switches·are activated and then P29 on that chip goes high and it starts all over again. 4) As to a link for learning about Machine intelligence.·There is a bunch, but I have found that a good book is better! Go to amazon.com type in "Technology" or "AI", (although I dislike the term), and you will find hundreds of texts on the subject.· Start out looking for ones that have basic learning in mind. The covers will often read something to the effect of: "AI the new frontier", sub title, "for the novice". Well happy learning to all, IAI Captain
P.S. If pressed I will try to find the exact article from Scientific American.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ IAI (Indigenous Alien Intelligence), The New View on Machine Intelligence. Because There is nothing Artificial about it!
That's a neat fact about a black hole being the limit of how big a computer memory can get. I didn't see that article but I did see one in another issue of Scientific American in which astronomers were puzzled over how black holes seemed to (according to some theoretical work) contain more "information" than the stars that collapse to form them, i.e. unexplained greater entropy. I think in the article they did resolve it, but it is also interesting that you really can think of a black hole as a massive "information storage". (Everything that it sucks up ends up squished into the surface but retains a pattern of what it was before it was squished; here "data compression" takes on a literal meaning - even the atoms are squashed flat!) I like the paradoxical way that a black hole can be considered a memory that you could write infinite amounts of data to, except that you can't get it back out. Sort of a no free lunch situation.
There is in fact a fascinating link between physical entropy and information capacity. Some good things to read in Wikipedia are about Maxell's demon, and Shannon's information theory.
Often entropy is a "bad" thing in a system, but it also reflects how much information is stored in the system. For instance, we store information on a CD by burning pits in it's surface. Compare an old, banged up tricycle to a new one at the store of the same model. The dings and dents in the old, beat up tricycle are entropy (the disorder increasing), but they also store information. Each mark and dent made is a record of the event that caused it. That's the analogical way I think of the relationship between abstract information and real physical entropy: you have to cause physical entropy - make some sort of mark on something in a way that increases it's physical disorder, in order to store abstract information.
The subject is fascinating because generally computer science issues seem abstract but every so often information theory turns out to have concrete impact on real physical things.
Supercomputing clusters was the first thing I thought of when I discovered the propller. I experimented with clusters some years ago using DOS and UBASIC. The nodes comunicated via serial and some of the techniques may apply to the propeller.
Not to be picky, but stacking propellers would made a supercomputing cluster rather than a supercomputer.
Comments
The patent system is to innovation what politcal correctness is to free thinking - death! To completely abolish the US patent system would be to give the real inventors in this country a chance to compete, long term, with the rest if the world.
I actually·just emailed EE Times earlier this week to·suggest they give my free subscription to some imminently qualified IP attorney, instead, since they seem to think these people·are the future. I·imagine (or like to imagine) that they got a lot of backlash from readers who complained about their March 20 issue where they glorified patent trolls as a 'New breed of innovator'. And, in their Crosstalk section where people write in, they published two letters, obviously from attorneys, who said that the US patent system is just fine after 200 years, and it shouldn't be changed. Right. This article you pointed out from April 3 was probably some compensatory reaction.·They've been leaning further and further left, anyway.·They emailed me back and said·"We are sorry you are so displeased with our publication." I am, too. I used to love it and recommend it to everybody. Here's the article, by the way. It might be good to be sitting down,·perhaps on the toilet:
http://www.eetimes.com/news/latest/business/showArticle.jhtml?articleID=181504068
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/21/2006 5:58:10 AM GMT
-Parsko
If it's doable, it would be pretty neat to develop a microkernel operating system where each process was run on an individual cog, or handled by islands of cogs.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
I'm glad you are into this parallel processing stuff, because I find it intriguing, too. If you want a 'sponsor' to try this out we could volunteer some chips. We don't have $40 million, though, or·maybe $100 million in today's dollars.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
(Is the die size a secret, by the way?)
Steve
-Parsko
The die size is 53 square millimeters - HUGE. If we hadn't used a full-custom approach, this would not have even been practical·in a·0.35um process.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/21/2006 5:25:45 PM GMT
http://research.microsoft.com/research/pubs/view.aspx?msr_tr_id=MSR-TR-2005-183
I bet you work for ASML.
We are using austriamicrosystems' .35um double-poly CMOS process. They were the only foundry we encountered·that·was eager to work with newbies like Parallax. It has worked out really well, though. Their quality has been excellent and it's been a pleasure to work with them. They are at least as much human as they are machine.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
By having X propellers enumerated each time they power up, will they enumerate to the same position each time? Or do you know that answer.
As for patents, I concur with both your views and feelings. But, I also believe that trade secrets have more discrete value than patents anyway. Last November the USPTO granted a patent to a man who had be filing since 1988. I was for the thought process to pay a manager. No formula, nothing other than a thought. It was reported in a UK programming journal in January. The patent system in the US has reached a point where it has so little value, it's not worth talking about. What lawyers haven't destroyed politics has. But I think it's going to get worse before it gets better. Rolling blackouts in Texas this week and people in your fair state are putting their jewelry in hock to buy gas. Hardly signs of a progressive society. But the real question here is, how expensive will gasoline get before the politicians start admitting how bad things are?
You are mentioning 2nd generation Propeller chips. Are they already on the drawing board? Also, what is the fastest clock input you have had the chip working with. I havn't seen the chip but it appears that it would handle more speed.
Congrats on you new product. I sent you an email weeks ago but you were probably too busy to see it. I'm looking forward to writing about it.
Oliver
Chip Gracey
Parallax, Inc.
Post Edited (Chip Gracey) : 4/22/2006 8:53:38 AM GMT
crumbs! That's a lot of silicon to be generating so little heat. Nice going [noparse]:)[/noparse]
>The problem is yield. Only ~84% of the dice are good
That's always been part of the fun of waferscale stuff, though - skipping the bad'uns, but still having the array be useful. (for some value of 'useful'...)
Still pondering what my killer app for Propeller is going to be. Too many choices - curse you and your excessively capable silicon [noparse]:)[/noparse]
Steve
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chip Gracey
Parallax, Inc.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Truly Understand the Fundamentals and the Path will be so much easier...
I delved back into reading about FRAMS microcontroller support chip which includes a 64 bit 'serial number' feature. It seems that it might be used to decode an encrypted software scheme.
Thus, you might protect future versions of your IDE from being used with other people's copies of your work.
I know it is a lot of might and maybe. I just don't have the know-how to achieve it myself.
Sadly, I suppose it would mean having a registered chip owner with a registered IDE.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"When all think alike, no one is thinking very much.' - Walter Lippmann (1889-1974)
······································································ Warm regards,····· G. Herzog [noparse][[/noparse]·黃鶴 ]·in Taiwan
The Spin interpreter is hardcoded for the Propeller machinecode, and any attempt of adapting it on another processor will demand a complete rewrite.
Besides, the cloners who have copied the Basic Stamps have a lot of catching up to do if they want to replicate the Propeller, and THAT's the clue, without a multicore chip(like the Propeller or the Cell -- which is supposedly used in the mythical PS3 -- processors) there's no reason to even bother looking at Spin.
Lets face it, the big ones, like Intel and Amd won't be interested in something like the Propeller(Intel, at least, have their own microcontrollers, running with high enough clocks to get the same MIPS out of them to rival the Propeller, and they've always thought that speed is king) and the smaller ones may not want to invest that much into a new architecture before they know it's a viable market, and even if they decided it is, they'd still have to develop their own chip, and as Parallax has proved, that takes time.
Even if they could assign a large group of people, I doubt they could do it in much less than two years, unless they want to end up with a very buggy and unstable product...
and by that time, we'll be ranting and playing with the next generation of Propellers...
The problem with serial-number ICs and encrypted software is that the Propeller doesn't have a protected ROM, so any decryption-software will be available for anyone to copy out and reverse.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Don't visit my new website...
A Few Pros
* With arrays of propeller chips, you could create neural networks.
* Propeller chips do not have operating systems - that is a big advantage over people running code on a pc with a windows xp home os and a P4 2.8Ghz processor (me.)
A Few Cons
* As someone I forget has said: multi-processor systems sum up to less than the·sum of the individual parts alone, because of time·spent·communicating, verifying, etc.
* These MCUs are still, (even in bunches) MCUs. There are other limits (like, is there hardware 32x32 multiplicaton or division?)
Conclusion
I don't think that multi-propeller boards will be seriously implemented, except for a few esoteric projects. It's still fun to think about, though.
Its simple enough for someone to dream of the need for lots and lots of pins.·
Also, it’s simple enough to use clever programming and hardware to provide other "nice" features between more then one controller.
The bottom line is how much effort does the designer want to go through to make a project happen?
For example:· I have no qualms about dedicating a propeller, and some hardware to control some servos, making this a subsystem within my project.
IN essence, this is the same as making an array.· Only, I'm specializing the functions of the subsystems, or atoms.·
Very similar to the human brain; the parts of the human brain that control speech cannot do anything with vision.·
Multi processing is something our brains to all the time, but with specialized interconnections and functions.·
Making an array of propeller chips to perform a give task, a task that leads towards the needs of multi processing, could lead to better performance.·
I agree with most of what was posted here, BUT:· I also believe in specialized functionality.·
Generic routines for each and all propellers, but, each would have it's own reason (code/hardware/object) for being included in the array.·
Just imagine something with the need to control 1152 servos.·
Forget about the power requirements, and other "hard facts", just think about how you would control them.·
In essence, you would load the exact same code into each propeller that would run the servo controlling object,
but EACH propeller would have something unique to it's self, so the object is actually running ALL the servos knows which propeller to send a servo request to.· Array programming coupled WITH specialized functionality.· Knowing redundancy is something of a necessity in any closed system, there would be some duplication of specialized functions, but that's duplication, not replication.·
No, I don’t believe in the statement that all nodes can perform any and all other nodes functions.· It doesn’t provide for specialized functions.·
If something needs 1024 bits of floating-point math, make an object (either hardware and or software), and give it to it.·
But don’t expand your expenses to give this to all other controllers in the array.· This is more then just an environmental or ecnomical issue.
Just because we want every house phone in the USA to be able to talk to every other house phone in the USA, doesn’t mean everyone’s phone is connected to everyone else’s phone, and it doesn’t mean that everyone has to have the same phone.
·
Yes, the idea of Multi Propeller applications is exciting, and would be a worthwhile learning experience for all who take that path, but the truth of it is this:
Given that all propellers from the same·family are created equally, there can be no gains, unless specialized functions are used within each.
....
Thank you for reading.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Just tossing my two bits worth into the bit bucket
KK
·
Since individual cogs can dynamically pull in program instructions they could reprogram each other or respond to stimulus.
Imagine a massively multiplayer online role playing game where the nonplayer characters were represented by individual cogs.
Given enough memory space they could remember who they met and whether they were treated well or badly.
You could send them on quests which would be small subprograms that they could load and execute.
If it was done well it would make a nice Turing Test...
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
I wonder if this wire is hot...
I'm currently doing undergraduate research in AI at the University of Oklahoma. I actually have built physical platforms (they move but need electronics added for brains) for a trio of giant robots (today, 3, tomorrow, an army) for which a "mobile supercomputer" of an 8x8 or even an 8x8x8 cube of Propellers would actually be appropriate and feasible. The four motors together peak at 100 amps current drain from large lead-acid batteries. The platforms are built with only one design goal: to be able to take more punishment than anyone could possibly expect a robot to survive, and keep coming back - I drove the prototype through waist-high weeds, knocked small trees over, ran it over cliffs, etc. Each robot is strong enough to carry a grown man on it's back; I could add 100 pounds more batteries to them without harming the physical frame or bogging down the motors - it definitely has the room and battery power for 64 propellers. One thing you could do with 512 processors is learning using a Genetic Algorithm. There are also robot control architectures (subsumption, vector summing, etc.) which fit the propeller-cluster hardware better than they fit the singule processor computers they are usually implemented on. However, what I would use a propeller cluster for is machine vision.
I'm currently only an undergraduate, but I plan to use these robots in my future Ph.d. work. I have been funding the project with my own money, so right now I have to settle for 1 propeller in each robot. If I got outside funding for it, and if parallax made such a propeller cluster board, then I would probably put them in these robots. But that's a lot of if's.
His last design was called the 25X. It had 5x5 of his forth engines on it and he claimed pretty much unlimited MIPS.
Seems like the propeller was a smaller realization of this theoretical chip. I will say that spin is eminently more readable compared to forth.
Is the spin interpreter a threaded interpreted language? It would make sense to me, as a forth interpreter is very small. Where each command is made up of a series of commands until they get to primitives which would be snippets of machine code.
I don't know if a multiprocessor board would be useful (right now). The high speed design with so much fan out would be expensive to design. You know how much of a problem the delays were with 8 on a chip die. Just imagine the problems taking it to FR4. I would wait until you see designs with two props appearing in the hobby world before going down this route.
Chip:
Can I ask one question? Are atmel and microchip afraid of you yet? If they aren't then they should be.
Post Edited (LucidGuppy) : 7/27/2006 12:36:03 AM GMT
Great job. Needs better documentation, but, great job!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Dave Evartt
People don't hate computers, they just hate lousy programmers.
http://wehali.com
Tell us about the "Next Generation Propeller Chip" that you mention in your initial post. That sounds interesting too.
Mark
Just a few points to comment on:
1) A 16 dimensional hypercube is called a Tesseract: The image of wich·is a simple three-dimensional representation reduced one further dimension by being viewed on a flat screen.
2) With regards to the max density that you could logically make A Prop matrix either as individual ICs or as one Die is the same! It dos not even matter what material it is made out of. For example: Even if it were made of tissue paper the results will still be the same. ·Read Scientific American, (can't remember the date, 2004/2005), but it has been calculated and found to the amazement·of the scientific community, (If for no other reason than no one ever bothered to think about it.), that the absolute highest memories density that can be achieved is the same density that is just short of a Black hole. If the density becomes any greater then it becomes a black hole and there for all stored memories are inaccessible do to the extreme high gravity.
3) With the coast of prop chips now being lower: It is possible·to build a 4x4 matrix for $192.00
···· As for the interface of the chips: Kiss them just I2c them together on P30&31 when p29 goes high then two things happen in parallel.
a) a simple I2c/(DS1wire ID&/ switch that turns on the next Prop, or any prop in any order, this is done by sending out a code to that switch in an order determined by·the master Prop, (ID#01, it's the prop·with an eeprom connected on P30&31). So! P29 going high says to the master prop I’m done and ready to be loaded with my new activities.
b) When the Master prop sees that P29 on the current prop IDxx is High it then sends it·data to that chip. the other chips wont be looking @ that time and the next or all unloaded chip/s will just be sitting Idle till there individual switches·are activated and then P29 on that chip goes high and it starts all over again.
4) As to a link for learning about Machine intelligence.·There is a bunch, but I have found that a good book is better! Go to amazon.com type in "Technology" or "AI", (although I dislike the term), and you will find hundreds of texts on the subject.· Start out looking for ones that have basic learning in mind. The covers will often read something to the effect of: "AI the new frontier", sub title, "for the novice".
Well happy learning to all,
IAI Captain
P.S. If pressed I will try to find the exact article from Scientific American.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
IAI (Indigenous Alien Intelligence),
The New View on Machine Intelligence.
Because There is nothing Artificial about it!
There is in fact a fascinating link between physical entropy and information capacity. Some good things to read in Wikipedia are about Maxell's demon, and Shannon's information theory.
Often entropy is a "bad" thing in a system, but it also reflects how much information is stored in the system. For instance, we store information on a CD by burning pits in it's surface. Compare an old, banged up tricycle to a new one at the store of the same model. The dings and dents in the old, beat up tricycle are entropy (the disorder increasing), but they also store information. Each mark and dent made is a record of the event that caused it. That's the analogical way I think of the relationship between abstract information and real physical entropy: you have to cause physical entropy - make some sort of mark on something in a way that increases it's physical disorder, in order to store abstract information.
The subject is fascinating because generally computer science issues seem abstract but every so often information theory turns out to have concrete impact on real physical things.
Not to be picky, but stacking propellers would made a supercomputing cluster rather than a supercomputer.
The site is still up here: http://dosbeowulf.tripod.com/