The video capabilities of the Prop I are what have lent it to retro design. I'm pretty sure we'll see that niche expand given the expanded memory and upgraded video capabilities of the Prop II. I agree that retro-gaming/computing is too small a niche to support a company the size of Parallax/Parallax Semiconductor. It'll likely be those interested hobbyists who continue to press that area forward. And why not?
I love being able to use the video out to drive small color LCD monitors
instead of being stuck using boring and limited HD44780 20x4 devices.
You are out 10.00 for the HD44780 and only 15.00-25.00 more to jump to a
much more capable 4" color NTSC display. The tiny 4 and 7 inch displays
let me make simple tools like analyzers that are much easier for technicians
to use. I can display things like help screens and put lots of info on screen
at once...you can't do that with HD44780 devices.
As you mentioned, the prop can stand in for a PC in some applications.
At a previous job I made a really inexpensive device using a prop that
had VGA output that eliminated the need for a customized laptop/netbook.
The notebooks running XP would have been a nightmare to use at hundreds
of locations. The prop device with kbrd and monitor required zero maintenance.
BTW: I don't have the electronics skills to design hardware like analyzers
and remote diagnostic tools. I come up with the (sometimes quirky) ideas
and write the software... someone with hardware skills was always at hand
to do the hardware for me. I know just enough about hardware to have a feel
for what is possible...and that seems to work for me.
As was stated in another thread, I believe what this means is that the dev tools are written in GCC, not that a GCC compiler will produce Prop object code.
While I too believed that at first glance, the last comment from Ken does not seem to lead in that direction. I want to hear Ken's response.
Strangely, none of our high-volume apps are display or gaming-oriented. They're all using the Propeller in ways that a general-purpose microcontroller would have caused a single-threaded, interrupt-driven programming disaster. And aside from a RAM shortage they're often left with resources to spare.
I think a lot of people don't realize just how many applications there are like this, and how few products there are for efficiently implementing them. While amazing things have been done with the P1, it might be worth backing up and considering that P1 was probably envisioned as the engine to replace and surpass the SX for BS3. For anything where you might even consider a Basic Stamp, the P1 is a delicious pile of tasty overkill. It only starts to look inadequate when you try making it run a whole computer full of peripherals that need drivers, and the fact that it's been pushed as far as it has in that direction is really amazing.
Neither the P1 nor the P2 needs to get into a stat fight with any other particular processor; what they need to do, as the P1 has done, is stake out new and uncharted territory in the cost / capability / ease of use phase space. Show me another CPU even remotely as capable as P1 that can be run with so few external components. P1 will still be standing all alone in that place when P2 comes out; P2 will need two power supplies and won't run on microamps in a pinch. But the tradeoff for that is the extra power you sometimes find yourself wishing for when working with P1. It's a different compromise, but one still far off the beaten path of what other sources are offering. The word Parallax Semiconductor needs to get out is that if a Propeller will do your application, it will probably do it with less external support and design effort than anything else out there.
ARM chips are becoming so complicated that they are moving towards the PC space. Some new ARM processors have 13 stage pipelines for media decoding and etc. Never will Parallax catch up with this. Parallax will find a home focusing on providing a fast, lean, and powerful MCU for developers with small teams interesting in producing reliable systems in medium volume (1K - 100K).
[1]Show me another CPU even remotely as capable as P1 that can be run with [2]so few external components.
[1] I can think of several off the top of my head.
[2] If I recall correctly the P1 (P2) requires an external EEPROM for the program code which by the way adds to cost, lost PCB real estate and code security.
Anyway back to the actual topic, a decent amount of RAM i.e. 256K to 512K would make the thing sing.
"Strangely, none of our high-volume apps are display or gaming-oriented. "
I'm a little surprised by that, but I guess it makes sense.
When I got into the Prop, I desired both deterministic operation and a VGA display.
The deterministic operation is absolutely world class.
I think the VGA and TV output is great, but I can see it being too weak for consumer applications.
But, I think Prop2 will change things. With a bit of external memory, we'll have graphical 1080p output with more color depth and fancy GUI options...
As was stated in another thread, I believe what this means is that the dev tools are written in GCC, not that a GCC compiler will produce Prop object code.
The other thread mentioned Eclipse & GCC in the same sentence, and the only thing really that makes sense in that context is using Eclipse as the IDE, and GCC as the C compiler to output Propeller object code. AVR & ARM are already doing this with good results.
[1] I can think of several off the top of my head.
[2] If I recall correctly the P1 (P2) requires an external EEPROM for the program code which by the way adds to cost, lost PCB real estate and code security.
[1] Got any examples? The only ones I know of either present a fixed menu of I/O peripherals (PIC) or need a lot more external support.
[2] You can get by without the EEPROM if there's another controller to load the program, in which case you can just about run the Prop directly from 2 AA batteries with no other external components at all. I don't think any of the other low end embedded controllers that are that forgiving have nearly the I/O capability of the Prop. Anyway, that's a design compromise which gives the Prop a superior fab process for other purposes.
Funny in that sense that I've recently had to design a prop-based audio/video shield for an arduino (not actually a sheild, the whole thing lives on one pcb, but you get the idea as to what the interface looks like) and it ended up where the prop is doing all the work and the arduino is logically not connected to anything anymore, but it cannot be removed from the PCB because they want to say that the project is arduino based. Derp!
Frankly if the arduino's programming app could be made to talk to a prop you'd see a mass migration, you get 8x the chip for the same price...
Too late? The Propeller is gaining in popularity - and Prop 2 is a great complement to the Prop 1. Without a doubt, it depends on your requirements and other microcontrollers can be better suited for particular designs.
Here's my question to the forum: What is the value proposition of the Propeller? And does the Prop 2 further deliver on that value, or offer something different?
IMHO - the Propeller 2 seems to be doing a job of maintaining what I love about the Propeller. Some things would be nice (like more RAM), but they come at a price, and keeping the chip inexpensive is important.
If we believe the specs for the Prop II, then it is top dog. You have 8 cogs at 160MIPS each, easy access to SDRAM, 128KB of hub ram (witch could easily be treated as L2 cache), and the resources to easily make an XMM manager that at worste would reduce you to 6 COGS at 80MIPS, with easy access to MBs of shared Program/Data ram. What more could we possibly ask for?
128KB of RAM is 4 times what the Prop 1 has. This will open up a lot of applications that currently don't fit in the Prop 1. Spin programs can be 4 times as large. Larger C programs can be implemented without requiring external memory. I see the Prop 2 as a step forward to a device that will have even more capability later on. A Prop 3 could have many of the features that people feel are missing from the Prop 2. It could come out within a year or two of the Prop 2 if Parallax decided to position the business that way.
It's just the beginning of something truly wonderful. There are already many things that can be accomplished with the Prop 1 that are spectacular. The Prop 2 will again outstrip our wildest imaginations in terms of new performances. I can envision so many incredible apps, that like the continuing solid BASIC Stamp's progression of varied and useful niche processors, the Propeller has a bright future!
I vaguely recall a Parallax person saying that the Prop2 would be ~the same cost as Prop1. Not sure I believe that though...
Hey Ray - we really don't know our manufacturing costs, so you have good reason not to believe that the price would be the same. Packaging and testing will be more expensive, but we'll only have one package [QFP] to stock which increases inventory turns.
So Parallax can kill Prop2 and ParallaxSemi and we can all be forthers happily ever after? You're so funny.
I think you're unfairly putting words into Prof Braino's mouth. He's just stating that the current propeller is plenty for his needs. I don't hear him trying to convince everyone that Forth is the One True Way nor that the Prop 2 should be abandoned.
Personally I am looking forward to Prop II becouse the increased speed and IO pins will let me use parallel memory at high speed to increase the data acquisition speed for a system that I am currently doing on Prop I.
Ken
Will Prop II be available on a prototyping board since I'm all thumbs mounting SMT?
LOL! So true! When I was actively involved with vision systems, I worked with QC departments a lot. Their mantra was always, "You can't inspect the quality in!" I guess the same is true for software. Following good software engineering practices alleviates the need for a lot of debugging.
My experience has been that when debugging tools were available and I leaned on them, my skills as a programmer did not improve. So I quit using them. What has helped my programming more than anything else is a philosophy of building and testing one step at a time. Constant testing is the key. What little debugging is necessary with each step can be accomplished with a serial terminal and an oscilloscope.
The biggest mistake a programmer can make is to write an entire program before testing it. Then -- yeah -- he better have good debugging tools and an extended deadline.
I think the 128Kb is just fine. More would be nice, but there are a lot of I/O pins. Much can be done.
I don't think any plans to break the symmetry of the COGs makes any sense at all. A COG needs to just be a COG. The moment that's not true, then code independence goes away, and with that a lot of the value of the Prop, IMHO.
For those, "I want one big cog" ideas, it's going to be possible to run very fast LMM code, and or XMM code, and there you go. Before breaking the symmetry in the chip, I think it would make sense to build the prop with external RAM as the hub, thus making it a real CPU, not a micro. (should be on the table for Prop III, IMHO)
As for retro type stuff. That's always possible on a good chip, and it's not the focus of this one, nor was it the focus of the last one either. No worries there.
Finally, the multi-prop design that Baggers and Coley are completing right now, demonstrates one element of the Prop that isn't fully exploited yet, and that is real parallelism. A few props, each with that 128Kb would be very potent, just as a coupla prop I chips are, and with the added I/O capability, many things will be possible that are almost there now, but for lack of pins and overall speed.
I worry about the focus on external memory, and or breaking parallelism and code independence. Those are the two primary attributes.
Bigger COG spaces make less sense to me as both the instruction set and overall speed ramp up. New instructions in the COG will pack a lot more in there from what I can see, and the higher speed of the Prop II means LMM type code is going to run seriously good. That's not getting the attention it should right now, and IMHO, that's lack of solid tools and such to make use of the technique.
Once we have more of that done, a Prop II can and will be able to run a lot of code, and do it threaded, out of it's HUB memory at speeds that exceed native on Prop I. Think about that for a moment, and other chips, and I think we are all not fully valuing all that Chip is doing.
That said, bring it on! Cheers to the faster releases too.
We are rapidly entering fun and very interesting times.
I have a commercial project that uses 3 props.
One prop could be replaced with a cheaper chip but not Prop II because it is located separately. But the cost consideration is simply another part, a new language, time to do the job. Simple, the prop wins hands down in R&D.
Two other props could be replaced with a Prop II BUT I would still require the 512KB of external SRAM because we are running Catalina and require big SRAM buffers for sorts, etc. Also, when the extensions to the software are complete, 128KB will not be enough. I am not interested in caching - adds to the complexity and I am damn sure I don't want to find bugs in that. (reminds me of the bloatware we have on pcs). Video is not used in this design.
So, C is a requirement although I could actually do the same in spin with LMM pasm helper code for the sorting, etc.
I think PropII will use video a lot more because I see a lot of niche products requiring displays - like 4.5", 7" & 10". We are seeing a push to HD in Prop II. This will require external SDRAM and this seems to have been addressed.
So we are back to 128KB of hub ram for code or else we have external SRAM too. Now we are going to be short of pins if we have external SRAM & SDRAM.
And it is true, we are only just starting to get the prop I to do what it is capable of. Prop II will be capable of a lot more and IMHO it will be constrained much more by the 128KB than anything else.
Ken, please, please get Chip to squeeze at least 256KB hub in there.
I think you're unfairly putting words into Prof Braino's mouth. He's just stating that the current propeller is plenty for his needs. I don't hear him trying to convince everyone that Forth is the One True Way nor that the Prop 2 should be abandoned.
You're kidding right? He is the great Forth evangelist.
He's even admitted to pushing it on some kids ....
I guess everybody has their own favorite subjects.
Beau will hopefully post a picture here soon of the current die.
It is already as big as the package it's going into allows. There is a huge DAC bus going around the interior of the I/O pads with all the memories located within that ring. What's left in the middle, you'll see, is where the million gates of synthesized mish-mash have to go. You'll see four big RAMs that are 32KB each. We need to implement them in sets of four because of the architecture. There is no room for another four of those blocks, so 128KB is all we are going to get into this TQFP-128 package using a 180nm process.
We are anxiously waiting to hear from the synthesis guys if the logic will even fit into the open space reserved for it.
There are 4,595 ports coming out of the synthesized block that must attach to the memories and I/O circuits. The synthesis job will probably take 2-3 weeks, as opposed to what was shaping up to be 2-3 years of schematic and layout work.
He's even admitted to pushing it on some kids ....
Wait a minute, Prof B is a FORTH pusher? Does that mean he has a FORTH lab somewhere near here full of solvents that might explode? Do we need to get the First Lady to give a speech telling our kids to Just Say No to FORTH? Are people growing FORTH under artificial lights in indoor hydroponic gardens? Are the Taliban selling us FORTH from their RPN gardens in Afghanistan? Are the Colombian FORTH cartels building submarines to sneak it past our C++ barriers? Is the FEA (Forth Enforcement Administration) gonna kick in my door one day because I read one of Prof B's posts and put me on a FORTH offender list for the rest of my life? Don't hold back man, we all need to know!
Comments
I love being able to use the video out to drive small color LCD monitors
instead of being stuck using boring and limited HD44780 20x4 devices.
You are out 10.00 for the HD44780 and only 15.00-25.00 more to jump to a
much more capable 4" color NTSC display. The tiny 4 and 7 inch displays
let me make simple tools like analyzers that are much easier for technicians
to use. I can display things like help screens and put lots of info on screen
at once...you can't do that with HD44780 devices.
As you mentioned, the prop can stand in for a PC in some applications.
At a previous job I made a really inexpensive device using a prop that
had VGA output that eliminated the need for a customized laptop/netbook.
The notebooks running XP would have been a nightmare to use at hundreds
of locations. The prop device with kbrd and monitor required zero maintenance.
BTW: I don't have the electronics skills to design hardware like analyzers
and remote diagnostic tools. I come up with the (sometimes quirky) ideas
and write the software... someone with hardware skills was always at hand
to do the hardware for me. I know just enough about hardware to have a feel
for what is possible...and that seems to work for me.
So Parallax can kill Prop2 and ParallaxSemi and we can all be forthers happily ever after? You're so funny.
I think a lot of people don't realize just how many applications there are like this, and how few products there are for efficiently implementing them. While amazing things have been done with the P1, it might be worth backing up and considering that P1 was probably envisioned as the engine to replace and surpass the SX for BS3. For anything where you might even consider a Basic Stamp, the P1 is a delicious pile of tasty overkill. It only starts to look inadequate when you try making it run a whole computer full of peripherals that need drivers, and the fact that it's been pushed as far as it has in that direction is really amazing.
Neither the P1 nor the P2 needs to get into a stat fight with any other particular processor; what they need to do, as the P1 has done, is stake out new and uncharted territory in the cost / capability / ease of use phase space. Show me another CPU even remotely as capable as P1 that can be run with so few external components. P1 will still be standing all alone in that place when P2 comes out; P2 will need two power supplies and won't run on microamps in a pinch. But the tradeoff for that is the extra power you sometimes find yourself wishing for when working with P1. It's a different compromise, but one still far off the beaten path of what other sources are offering. The word Parallax Semiconductor needs to get out is that if a Propeller will do your application, it will probably do it with less external support and design effort than anything else out there.
[1] I can think of several off the top of my head.
[2] If I recall correctly the P1 (P2) requires an external EEPROM for the program code which by the way adds to cost, lost PCB real estate and code security.
Anyway back to the actual topic, a decent amount of RAM i.e. 256K to 512K would make the thing sing.
I'm a little surprised by that, but I guess it makes sense.
When I got into the Prop, I desired both deterministic operation and a VGA display.
The deterministic operation is absolutely world class.
I think the VGA and TV output is great, but I can see it being too weak for consumer applications.
But, I think Prop2 will change things. With a bit of external memory, we'll have graphical 1080p output with more color depth and fancy GUI options...
The other thread mentioned Eclipse & GCC in the same sentence, and the only thing really that makes sense in that context is using Eclipse as the IDE, and GCC as the C compiler to output Propeller object code. AVR & ARM are already doing this with good results.
[1] Got any examples? The only ones I know of either present a fixed menu of I/O peripherals (PIC) or need a lot more external support.
[2] You can get by without the EEPROM if there's another controller to load the program, in which case you can just about run the Prop directly from 2 AA batteries with no other external components at all. I don't think any of the other low end embedded controllers that are that forgiving have nearly the I/O capability of the Prop. Anyway, that's a design compromise which gives the Prop a superior fab process for other purposes.
Funny in that sense that I've recently had to design a prop-based audio/video shield for an arduino (not actually a sheild, the whole thing lives on one pcb, but you get the idea as to what the interface looks like) and it ended up where the prop is doing all the work and the arduino is logically not connected to anything anymore, but it cannot be removed from the PCB because they want to say that the project is arduino based. Derp!
Frankly if the arduino's programming app could be made to talk to a prop you'd see a mass migration, you get 8x the chip for the same price...
Here's my question to the forum: What is the value proposition of the Propeller? And does the Prop 2 further deliver on that value, or offer something different?
IMHO - the Propeller 2 seems to be doing a job of maintaining what I love about the Propeller. Some things would be nice (like more RAM), but they come at a price, and keeping the chip inexpensive is important.
This is a real factor IMO.
Hey Ray - we really don't know our manufacturing costs, so you have good reason not to believe that the price would be the same. Packaging and testing will be more expensive, but we'll only have one package [QFP] to stock which increases inventory turns.
Ken Gracey
The web site should address the debug issue. I'll mention it on the other thread.
Ken
Will Prop II be available on a prototyping board since I'm all thumbs mounting SMT?
-Phil
The biggest mistake a programmer can make is to write an entire program before testing it. Then -- yeah -- he better have good debugging tools and an extended deadline.
-Phil
I don't think any plans to break the symmetry of the COGs makes any sense at all. A COG needs to just be a COG. The moment that's not true, then code independence goes away, and with that a lot of the value of the Prop, IMHO.
For those, "I want one big cog" ideas, it's going to be possible to run very fast LMM code, and or XMM code, and there you go. Before breaking the symmetry in the chip, I think it would make sense to build the prop with external RAM as the hub, thus making it a real CPU, not a micro. (should be on the table for Prop III, IMHO)
As for retro type stuff. That's always possible on a good chip, and it's not the focus of this one, nor was it the focus of the last one either. No worries there.
Finally, the multi-prop design that Baggers and Coley are completing right now, demonstrates one element of the Prop that isn't fully exploited yet, and that is real parallelism. A few props, each with that 128Kb would be very potent, just as a coupla prop I chips are, and with the added I/O capability, many things will be possible that are almost there now, but for lack of pins and overall speed.
I worry about the focus on external memory, and or breaking parallelism and code independence. Those are the two primary attributes.
Bigger COG spaces make less sense to me as both the instruction set and overall speed ramp up. New instructions in the COG will pack a lot more in there from what I can see, and the higher speed of the Prop II means LMM type code is going to run seriously good. That's not getting the attention it should right now, and IMHO, that's lack of solid tools and such to make use of the technique.
Once we have more of that done, a Prop II can and will be able to run a lot of code, and do it threaded, out of it's HUB memory at speeds that exceed native on Prop I. Think about that for a moment, and other chips, and I think we are all not fully valuing all that Chip is doing.
That said, bring it on! Cheers to the faster releases too.
We are rapidly entering fun and very interesting times.
I have a commercial project that uses 3 props.
One prop could be replaced with a cheaper chip but not Prop II because it is located separately. But the cost consideration is simply another part, a new language, time to do the job. Simple, the prop wins hands down in R&D.
Two other props could be replaced with a Prop II BUT I would still require the 512KB of external SRAM because we are running Catalina and require big SRAM buffers for sorts, etc. Also, when the extensions to the software are complete, 128KB will not be enough. I am not interested in caching - adds to the complexity and I am damn sure I don't want to find bugs in that. (reminds me of the bloatware we have on pcs). Video is not used in this design.
So, C is a requirement although I could actually do the same in spin with LMM pasm helper code for the sorting, etc.
I think PropII will use video a lot more because I see a lot of niche products requiring displays - like 4.5", 7" & 10". We are seeing a push to HD in Prop II. This will require external SDRAM and this seems to have been addressed.
So we are back to 128KB of hub ram for code or else we have external SRAM too. Now we are going to be short of pins if we have external SRAM & SDRAM.
And it is true, we are only just starting to get the prop I to do what it is capable of. Prop II will be capable of a lot more and IMHO it will be constrained much more by the 128KB than anything else.
Ken, please, please get Chip to squeeze at least 256KB hub in there.
He's even admitted to pushing it on some kids ....
I guess everybody has their own favorite subjects.
It is already as big as the package it's going into allows. There is a huge DAC bus going around the interior of the I/O pads with all the memories located within that ring. What's left in the middle, you'll see, is where the million gates of synthesized mish-mash have to go. You'll see four big RAMs that are 32KB each. We need to implement them in sets of four because of the architecture. There is no room for another four of those blocks, so 128KB is all we are going to get into this TQFP-128 package using a 180nm process.
We are anxiously waiting to hear from the synthesis guys if the logic will even fit into the open space reserved for it.
There are 4,595 ports coming out of the synthesized block that must attach to the memories and I/O circuits. The synthesis job will probably take 2-3 weeks, as opposed to what was shaping up to be 2-3 years of schematic and layout work.
May the FORTH be with you!