Licensing the Propeller 2 FPGA: For Serious Commentary and Consideration
Oldbitcollector (Jeff)
Posts: 8,091
I mentioned this idea in another Propeller 2 thread, but I have since retracted all of my recent threads as they were too critical of Parallax's operation. I've received the "go ahead" that this concept is one is solid for public consumption, so I'm going to put on my very best word-smith cap and open this discussion hopefully without invoking hard feelings and concerns...
<deep breath> here goes...
(First a little history from my favorite era and subject. Stick with me here, you youngsters might learn something.)
Back when Apple was still producing machines that displayed "]" prompts and booted with something called PRODOS, a young Steve Jobs got a chance to "time travel" in a sense to the future with a visit to Xerox Parc, a research division where Xerox had hired the top minds to "create the future". They had! They had invented Ethernet, Email, and most importantly the GUI (Graphic User Interface). Xerox seeing themselves as a company who put things on paper didn't see the value in the creations from their own research company, but Steve Jobs KNEW that this was the future of computing and after a visit, returned the next day with several Apple engineers armed with notebooks. The end result was the Macintosh Computer which looked very strangely like the Xerox Altos prototype computers they had seen.
Fast forward just a couple short years.. Another young programmer by the name of William Gates got a look at the Macintosh and realized that the GUI had more to do with the PROGRAMMING CODE than the machine it was running on. He ran with the idea of moving computers from a C:/> prompt to something that we know as Windows today. The magic was that the GUI is simply software. Guess what? It's software on the modern Macintosh too.
Parallax has been pushing for a long time to get the Propeller 2 silicon (hardware) to market. They have suffered from redesigns caused by technology changes, on top of this there is much hardware testing that needs to be done to prove the chip.
Here's the crazy idea: Hopefully better presented...
License the Propeller 2 FPGA as SOFTWARE instead of pushing to silicon. Make the license free for personal/hobby use (registration required) because these are the people who will spend countless hours exploring, pushing, writing code, blogs, and documenting their findings. Discount the licensing for educational use. (I know that Parallax already offers a better price on hardware when it's purchased at quantity 100 or 1000. The same would happen here.) License the code for commercial use. Again with a provision for a slightly better price for those who purchase 1000 licences at a time vs the small hardware producer who buys a few at a time.
Here's are the neat kickers. First, Chip could continue to develop the code and make improvements. (Registration over piracy has it's privileges with upgrades and support!!) Parallax could create some great FPGA products who run the code on. If created cooperatively with a few other vendors and "insiders" the boards could be tasked to do all kinds of great things. You would be surprised all the places that the Raspberry Pi boards are being found these days. This would be no different.
Finally, we could start working NOW. We still need some real development time to get traction under code. The Propeller 2 will be no different from the Propeller 1 in this regard.
Again, this is IN NO WAY intended as a criticism of the decisions of Parallax, it's simply an idea from someone who still wants to "live the dream" that's been a long time coming.
Jeff
<deep breath> here goes...
(First a little history from my favorite era and subject. Stick with me here, you youngsters might learn something.)
Back when Apple was still producing machines that displayed "]" prompts and booted with something called PRODOS, a young Steve Jobs got a chance to "time travel" in a sense to the future with a visit to Xerox Parc, a research division where Xerox had hired the top minds to "create the future". They had! They had invented Ethernet, Email, and most importantly the GUI (Graphic User Interface). Xerox seeing themselves as a company who put things on paper didn't see the value in the creations from their own research company, but Steve Jobs KNEW that this was the future of computing and after a visit, returned the next day with several Apple engineers armed with notebooks. The end result was the Macintosh Computer which looked very strangely like the Xerox Altos prototype computers they had seen.
Fast forward just a couple short years.. Another young programmer by the name of William Gates got a look at the Macintosh and realized that the GUI had more to do with the PROGRAMMING CODE than the machine it was running on. He ran with the idea of moving computers from a C:/> prompt to something that we know as Windows today. The magic was that the GUI is simply software. Guess what? It's software on the modern Macintosh too.
Parallax has been pushing for a long time to get the Propeller 2 silicon (hardware) to market. They have suffered from redesigns caused by technology changes, on top of this there is much hardware testing that needs to be done to prove the chip.
Here's the crazy idea: Hopefully better presented...
License the Propeller 2 FPGA as SOFTWARE instead of pushing to silicon. Make the license free for personal/hobby use (registration required) because these are the people who will spend countless hours exploring, pushing, writing code, blogs, and documenting their findings. Discount the licensing for educational use. (I know that Parallax already offers a better price on hardware when it's purchased at quantity 100 or 1000. The same would happen here.) License the code for commercial use. Again with a provision for a slightly better price for those who purchase 1000 licences at a time vs the small hardware producer who buys a few at a time.
Here's are the neat kickers. First, Chip could continue to develop the code and make improvements. (Registration over piracy has it's privileges with upgrades and support!!) Parallax could create some great FPGA products who run the code on. If created cooperatively with a few other vendors and "insiders" the boards could be tasked to do all kinds of great things. You would be surprised all the places that the Raspberry Pi boards are being found these days. This would be no different.
Finally, we could start working NOW. We still need some real development time to get traction under code. The Propeller 2 will be no different from the Propeller 1 in this regard.
Again, this is IN NO WAY intended as a criticism of the decisions of Parallax, it's simply an idea from someone who still wants to "live the dream" that's been a long time coming.
Jeff
Comments
Here's the crazy idea: Hopefully better presented...
License the Propeller 2 FPGA as SOFTWARE instead of pushing to silicon.
These two are not mutually exclusive.
Finally, we could start working NOW. We still need some real development time to get traction under code. The Propeller 2 will be no different from the Propeller 1 in this regard.
Err, There is no need to license the core, to have testing done.
Older P2 images have already been tested on FPGA boards.Newer P2 images do not yet exist, so again, license does nothing magical to get those sooner,
What has happened during the P2 gestation(s), is the FPGA's continue on their price curve, and the Free Software Device support threshold has increased.A $150 Board, and Quartus WebPack now supports the A9 CV.The MAX 10 is also a candidate, and supports P1V now. (and also supports Altera NIOS)
However, even with these movements, the FPGA Price points are still way above MCU prices, which underlines that Silicon and FPGA paths are not mutually exclusive.There is a quite separate, and large, market for Silicon vs FPGA alone.
Given the sunk costs of Silicon development, I cannot see Parallax not moving to Silicon.FPGA image licenses are a separate, independent, question.
As a quick reality check, look at this news item, and count the cores/IP in this 16nm tapeout
http://forums.xilinx.com/t5/Xcell-Daily-Blog/We-have-Tapeout-Xilinx-Zynq-UltraScale-MPSoC-ready-for-TSMC-s/ba-p/639618
Of course, P2 could tile into the Logic fabric of that device.
To pull it off, Parallax have to make the case that a PII soft core on a FPGA is superior to the current soft cores out there in terms of performance, flexibility and cost-effectiveness, etc.
Here's the rub, it's cost. The fact that a PII can be stuffed in the latest offering from Altera isn't a selling point, it's a liability because of cost.
Why should a company shell out a license for a soft-core that can only run on the latest FPGA monster that costs $200+ for singles? When they can go out and buy a dual core Delfino or Freescale dual core PPC real time controller for $30 in singles or simply use a soft-core that that can fit in a $20 FPGA?
Why should the costs be a factor at this point?
1) I suspect that Parallax has the ability to wear the big pants and purchase a quantity of chips as needed. In this case FPGA.
2) We still have at least 1-2 years of maturity time in which this new design "code" will require the love and care from all parties involved. I'm positive that the costs of FPGA technology will continue to go downward in this time.
3) I suspect that Chip can make it manageable inside a reasonable FPGA package. For all that matter if it comes down to code size, why not offer it in more than one flavor?
I know, it's a crazy idea... But sometimes someone makes a crazy idea that will cause decision makers to say "No, that's not really possible, but you know we could do.............."
Look at the ARM recipe. That certainly works! You can buy a simple design, or the "whole enchilada" like Apple did.
And it could help fund the real silicon P2.
Remember, a number of users are not price conscious like the hobbyists. And you can use a smaller FPGA if you don't require all the features.
The "hot" P2 was running, so there are plenty of P2 versions/variants possible.
Ken even knows who the first customer would be
I don't get it. Are you suggesting that there never be an actual PII chip from Parallax? That I need a big honking expensive FPGA to use the PII HDL IP with? Or I have to hope that some chip vendor licences the IP and makes PII chips.
Sounds expensive and/or complex and makes getting back into XMOS devices very attractive.
By the way we can't say that Parc invented the GUI after seeing the "Mother of all Demos"
I'm sure the young Bill Gates looked at the Mac and thought "We had better get one of those GUI things else we are out of business"
Bill's big realization was that little computers were going to be everywhere and without software they are pretty useless. Enthusiastic pioneer buyers of Altairs and the like needed help.
One could say that FPGA's are a similar blank slate as those software deprived early computers. Ready to be filled with PII images. Except I don't see the world storming into the FPGA idea and being ready for someone to fill that need. Too big, too expensive, too complex.
(It got way too hot in the sticks. Dogs broke down, and we've returned home early.)
1. Absolutely nothing should get in the way of real silicon. The Propellers have a long sales life. Getting P2 done means a long term revenue stream as well as an actualized device, both matter to us and Parallax, and both lend credence to a license scheme.
2. Licensing can happen apart from the silicon. This means considering it as it's own idea, using information from the current effort. I see few reasons why this is a problem.
3. What about the smart I/O? A big part of what will make a P2 do what it does is the nature of the custom pins. A licensed design, unless made in silicon ARM license style, won't really have the attributes intended in the P2 today.
3.1 Does this matter?
3.2 What if a small subset of smart I/O were mostly possible in an FPGA, or through some add on board much like we saw done last go around?
4. Both licenses and real silicon will need to center in on a reference, baseline implementation. They may diverge as licensees see fit, but there really does need to be some definition of what is a P2 and what is a P2 plus some stuff some people wanted or didn't want.
5. The work required to get to a license ready state is still in progress and it's not likely to get there, until we've got runnable FPGA and the initial round of testing / tools to go with that.
5.1 Chip has mentioned a self-hosting option. In this context of licenses, I believe that is a kick Smile feature. It can be one self-contained package. This does not mean other tools aren't going to be used, or even that they won't be the majority use case. But it does provide a very clear definition of what is a P2 and SPIN, etc...
5.2 Where it's not self-hosting, the current plan is to build that tool set together. GCC and friends on one axis, Chip and SPIN + PASM on another. There is going to be some outside of Parallax help and documentation happening. How to handle this in a license scenario?
6. I am opposed to variants at first, and I'm thinking specifically of P2 "hot" edition. That one is kind of awesome, but it's never going to see the light of day, and it's a very confusing and potentially distracting thing to have out there given the focus on a viable silicon design.
7. Licensing that tech targeted for real silicon provides a path for people who are developing products or solutions now, willing to pay for early adoption and presumably able to ask for appropriate margins. For those who are successful, a transition away from an FPGA to silicon will be very attractive. This seems the optimal path to me.
8. Why not just use the FPGA image?
8.1 Seems to me, being able to link a bunch of stuff together in an FPGA package would simulate being able to do the same with real silicon in a package, and this gets products cooking earlier than we would see otherwise. That's the most compelling reason for licensing, given dicipline on keeping this design viable and that Chip gets to follow through on his vision for the device.
8.2 This point is sensitive to me. I'm interested in seeing it self-host, and to that end, really want the monitor, assembler, SPIN, etc... components done by Chip to capture the benefit of how he sees programming differently from the vast majority. All of us may or may not be interested, and it may or may not mesh with our preferred processes and tools too. No matter. Having that play out as it did on P1 will be very highly differentiated and IMHO, compelling enough to bring users just as P1 did.
8.3 I'm in support of licensing and willing to help, given we don't break what I put in #8 here.
It's an idea worth discussion OBC. I'm in agreement over it being a slightly faster, though risky, path to doing real things with the design in FPGA.
You've got my initial thoughts. Now I'm going to find somewhere to cool off and catch up!
I agree. The P1V has not seen a flood of commercial applications, but it has educational merit.
The dual aspects of test coverage & Educational use are why I think Parallax chose to do their own FPGA board.
There is no rush needed to license P2 images, first P2 needs to work, & software infrastructure needs to be built.
Does anyone outside of Parallax know if this is true or not?
I'm only aware of the commercial applications I'm involved with or those pointed out in the forum. I have absolutely no idea how many Propellers are sold.
Was your comment a guess or do you have inside information?
Sorry jmg, I should have read more carefully. You said P1V. I imagine your statement is correct though I have no way be being sure if the P1V has commercial applications or not.
Sorry jmg, I should have read more carefully. You said P1V. I imagine your statement is correct though I have no way be being sure if the P1V has commercial applications or not.
I'm sure P1V commercial application are possible, but the chatter of those building P1V code has not yet indicated any commercial imperatives.
I think the MAX 10 is well suited to P1V, and the NIOS MHz numbers I quoted in another thread ( 270MHz II/f, 330MHz II/e) indicate MAX 10 should give a solid P1 speed boost.
Prices are now up on the whole MAX10 range, cheapest/smallest is $2.85/1k, and range up to $150+
Some build stats for MAX 10 P1V are here
http://forums.parallax.com/discussion/comment/1315110/#Comment_1315110
Looks just under 2K LE/Cog, making a real P1 cheaper per COG, but leaves niches where the higher MHz is mandated, or logic fabric can be used.
Wonderful video! Thank you for posting that. I still think Xerox did enough work to say that their name should easily be on the machines we are both using right now. Shame they just didn't have corporate vision to see what their own enginneers had created.
Unless the P2 itself shows some very sexy abilities that other parts don't/can't do at similar or lower cost, I fail to see where any sort of competitive advantage would come to play to push anyone of scale to get into the FPGA-P2 pool.
Worse, if P2 does ever show some sort of advantage, at 0.01mm/core M0+, or larger M3/4 could be multi-cored up to 16 or more by any of the players. Luckily for Parallax, they're all chasing the phone/auto markets at Ghz speeds.
Ultimately, I am not sure whether the P2 will be enough to do more than migrate its customer base from P1 to P2, and perhaps expand somewhat. Sort of like how the world was supposed to be Microchips oyster when they came out with the Pic32 MIPs.
Bit as long as it helps the revenue stream...
@Heater. Fabulous video!
As Xerox PARC was built on land leased from Stamford University and hired graduates from the same, one would assume that one of these graduates saw that demo in 1968 and mentioned it @PARC in the early 70's.
Parallax is the copyright holder of whatever IP goes into the P1V. As such they can distribute it how they like to whom they like with whatever licensing they like. The current license is what we agree to when we download from github. Nothing stops you approaching Parallax and asking for commercial closed licensing or whatever.
Yeah, I'm sure word of that demo got around. Not that I expect any of the stuff shown there was secret in any way.
Well, you are because communication is so hard here...
That should be a bit interesting.
Would the license make sense in that context? People would be using an FPGA anyway, but may well find a 16 COG device as a significantly easier path to develop on it.
Of course doing it FPGA and HDL only would cost less and fit into a smaller device at the cost of doing the HDL and time to market and or potentially having to include a CPU core as external device or HDL core anyway.
If we are sure this one is going to silicon, this niche may make sense for people wanting to produce products. The cost and margin on those products may easily tolerate the higher, early cost too. Depends on the business model and niche.
This is why I thought the idea worth discussion.
As for software, tools and such...
We know Chip will eventually make SPIN, PASM tools that self host. PASM will come easy and will happen as development does. The last image was favorable in that regard. PASM with HUBEXEC won't be a pain. Frankly, it will be easy. Many of us got to give those ideas a test drive.
A macro assembler won't be too long after that, nor will gcc.
A finalized SPIN will likely follow.
Anyone wanting to do something along these lines could build a lot with PASM and migrate to the more robust tools as they come online, as well as move to silicon when it is done.
...or maybe not. A P2 embedded in an FPGA along with some other things may well save big on board manufacturing costs.
Personally, I've some industrial targets in mind. P1 can do the simpler ones now. A P2 that I know won't end up non silicon, like the hot edition is, would be a great start. I'm not sure people realize how excellent building bigger things without an OS is going to be.
You're getting way ahead of where things are and need to be before there can be any talk about licensing the PII.
Wait a year or two and see how the silicon compares to the real time I/O controllers that Freescale and TI among others have out for automotive and industrial apps.
That should be a bit interesting.
Let me modify that a bit...
$1 USD to all buyers within the first 90 days, and then $250 USD thereafter to buyers who desire to provide their own FGPA. If you were to buy a FGPA with the bundled license from Parallax, the price of FGPA would include a license.
It is time for Parallax to start thinking about some sort of promotion and incentives for early adopters rather than sell of bit of this, sell a bit of that. The increase in sales and cash flow might just move the actual Propeller 2 into production a bit faster as demand would grow with this kind of approach.
http://www.xmos.com/products/silicon/xcore-200
It won't be long before they are in full production.
BTW, Freescale has merged with NXP.
http://www.xmos.com/products/silicon/xcore-200
It won't be long before they are in full production.
and then there is this, next-gen Zynq, from Xilinx..... http://forums.xilinx.com/t5/Xcell-Daily-Blog/We-have-Tapeout-Xilinx-Zynq-UltraScale-MPSoC-ready-for-TSMC-s/ba-p/639618
Of course a p1V or P2 could drop into the Logic of that if there were a compelling use-case.
You can compile other languages into an FPGA configuration besides Verilog or VHDL. What about SystemC? I think there are even compilers for C and probably other languages. Has anyone tried any of these as a way to easier use of an FPGA?
In what context is that comparison a qualifier?
Certainly not all of them. Someone familiar with a P2 may well have a compelling reason to build things on it in advance of silicon.
RE: Other languages. I didn't see anything there with the lean, easy nature of Propellers. That is a valid path though. More than one way to use an FPGA.
This way might be the "doesn't need to know much" way which may well put normally out of reach things into the hands of people able to make products.
And what of people who may well have an application in mind right now? I know what I could do with the last iteration. It is powerful and not all that hard to use.
Put another way, I could get something done in the time it would take to ramp up on other things. That is a compelling idea.
This one is extremely likely to be similar.
One other thing in the post Jeff made:
The design would see more than testing prior to silicon. Has anyone noticed how Chip can package complex things into some pretty easy, clean and robust ways?
Sunk costs are a financial fallacy on par with the gamblers fallacy. When doing novel things, R&D dollars do not always pay off. In this case, we have somebody with dreams of doing it differently. Either it happens, or it does not. There is no Wall Street board holding those dollars to account.
It is entirely rational to consider those costs a mere expense, compute what is possible today and try again.
The same tired cost arguments and feature comparisons are not factoring in how others may choose to do business. Jeff is right in that cost isn't a factor at this time.
Let those wanting to make stuff figure cost out. As mentioned, there are a ton of niches that can be coupled with subject expertise to be exploited. Margins are all about value perception.
BOM costs suggest a minimum floor, but there is always room to add value above that and deliver a nice product people will pay for even when it is not the cheapest possible product.
This isn't about getting established players to adopt the Propeller way. It is all about making the Propeller way accessible to people looking for ways they can use to get things they want to do, done.
Recently, I watched an industrial machine builder pay a lot of money for a servo with a little interface board attached. Seriously, the price is insane compared to a controller board and servo and some software.
Why?
"Oh, I can put a few buttons and dials and switches on a panel and be done."
That guy doesn't do programming and paid easily and paid a lot for something enabling in a way the did not require programming.
The P2 case is similar. People could do some pretty neat stuff that would often require a lot more of them should they be using more common, established tech.
Parallax is never going to be able to keep up with the competition with making chips. They do not have the resources.
They could however bring their vision to custom FPGA IP that could leverage other people's hardware to their advantage and profit.
Just look at Mesa. They sell a wide range of their AnythingIO modules all based on FPGA boards. That is where I have moved to with my projects. The boards cost way more than chips, but they are complete ready to plug and play I/O controllers. Not only that they are open source and customizable with free Xilinx tools.
What they lack is on board control capability (processor).
My direction is to use a larger FPGA board, with ARM cores, and add a custom processor (inspired by the P1V and RISC V) between their I/O logic and the ARMs to create a souped up version of the BeagleBone Black. I would LOVE to be able to go to Parallax and shop for a P1V+ or P2V variant to do what I want.
A series of boards based on the Altera Max10 series (as an example) could be made as drop in System-on-Modules that could be designed to be easy to adapt to other systems or uses. Complete flexibility! Education, hobby, custom commercial applications all pulling from the same investment.
Times change, and are changing even faster now. By the time Parallax gets a P2 chip in hand it will be obsolete with little to no real market potential.
Perfect example is their 123 board. Spent a bunch of time and money to
design and manufacture it and now there is an A9 based board for 1/3 the price already
to market while we still wait for theirs.
On the other hand Parallax and Chip could keep a constant stream of leading edge ideas/designs flowing out in soft IP that would keep them in the game. Look for new hardware, boards, applications and quickly demonstrate ways to put them to use with the brilliance that is Chip/Parallax.
ARM ARM ARM! Follow their model and leverage your genius!
One wakes up one day with are warehouse full of out-dated items that can only be sold to Hong Kong to be put on E-Bay at fire sale prices.
The biggest problem is that Parallax never is able to get the volume of sales that brings down costs via economy of scale unless they keep a very narrow product line.
What they do is target a couple sweet spot FPGA devices, produce reference design boards, like the 123, and build out the tools, education, etc... on those.
Every so often, the reference design gets refreshed or a new one created.
If one buys a board, a license is included. That way, people can combine Propeller 1 or 2 or a variant with other things and a few such combinations would be in the reference design to show the way and such.
Or, people can get a license and roll their own. Community support on that is the default.
If people want or need more, they use the reference materials, or they buy support from Parallax.
If Chip is up for it, buying additions to the P2 or variants may make a lot of sense and he's in a sweet spot and maybe funding another dream he has.... just a thought.
I personally think every license should cost a little something. Just enough to establish a relationship should other business get done and to focus efforts a little, bit that is just me.
@Kerry. Yeah, right on. I do remember those comments and agree mostly.
I disagree about real chips. Releasing a reference design in silicon will work about like P1 does and there is a nice business there when it's done. It won't change the market, but it will serve nicely for people doing things in a more simple way where an FPGA is too much. Lots of that going on and it will continue.
Western Design Center is still doing 6502, and it's a nice little business.
P1 and friends could be a nice little business too. No need to change the world.
It won't be about that. It will be all about value for the dollar.
Again, the guy I watched pay very large amounts of money for servos with some little board setup on them is a case in point. He does not program, but he does understand pots, switches, simple sensors, etc... that servo has built in functions for a lot of common stuff.
To him, worth every penny. Value for the dollar.
Most here would never, ever pay for that because we could build a better thing specifically.
A P2 on FPGA will present a similar kind of valentine proposition in that it could make some pretty advanced things accessible to people who would not do it otherwise, just as P1 has done.
Take a look at Parallax's HB-25 motor controller $50 (https://www.parallax.com/product/29144). If you have never had the chance to work with one of these, let me tell you they are worth every penny. I can purchase pretty much the same thing out of HK for around $11, but it won't be nearly the quality of the Parallax unit.
The same will go with the FPGA licensed P2. If done right, it could be a perfect solution even if it costs a little more.
I'm really not worried about the higher price of FPGA technology. It WILL continue to get cheaper going forward. It's already happening.