Not to reopen an old can of worms, but how are propeller enthusiasts getting along with this chip now that they have given it some of their time?
I saw this post on the XMOS forums (back in October) regarding several propeller enthusiasts active in these forums, apparently crossing over to XMOS: www.xlinkers.org/forum/viewtopic.php?f=3&t=93&st=0&sk=t&sd=a
Although Leon and Ale have developed some code for the XMOS and Leon is probably more active in the XMOS forums now than the propeller forums, several of the other folks (heater coley, timothy, cluso) mentioned in the thread above seem to have not gone much further with the XMOS. Was productivity trickier than the Propeller?
Their is a nice set of code examples accumulating in their obex: www.xlinkers.org/projlist
But it is still no where near the level that is in the Propeller Obex
I've been working with two other regular posters to these forums on a multi rotor craft and one recently decided the XMOS was a better solution for his efforts than the Propeller. We debated crossing over to the XMOS with him, but will stick with the propeller temporarily. I think all three of us are curious how the pioneers listed above have found the development experience with the XMOS chips relative to the Propeller though.
In a previous post by Leon he indicated there will a single cog version available in volume at $1. This does seem very interesting. For example I may want to use Propeller code for a task, but only need 1-3 cogs and can't justify a $15 chip, especially if considering volume. But I could not find this statement on their forums, so was not sure how that price was derived - apparently it will be released later this year.
Looking at the underside of the evaluation pcb posted by timothy, it seems there are many more component dependancies for an XMOS setup than Propeller. The other concern is that this requires a 6 to 8 layer pcb, and that takes me outside the realm of free versions of Eagle or Target3001. www.flickr.com/photos/brilldea/2963394963/sizes/l/
You can get a more complete dev kit too, the XDK, but are they serious? $999 - that's not hobbyist friendly
I'd like to ask how the Propeller II will match against this, since I might be less tempted to stray, but I know that's taboo.
There are a few applications where the XMOS would be a good choice but I shy away from the power (120ma quiescent) and pcb requirements (BGA multilayer) of a monster chip. A lot of applications can get by with a simple simple 8-bit CPU so Propellers eight 32-bit cores might seem overkill for most applications yet the Prop is in my opinion even easier to use than the 8-bit'rs. As for that magic $1 silicon I will believe it when I see it but I very much doubt that will happen.
Good luck to the XMOS boys and I may hop on that wagon if it's going my way but my love is the Prop chip. It is both a "hobbyist" chip and also a very serious (but fun) commercial chip. Viva la Prop!
Prop II will be here before we know it but a watched kettle never boils, so keep busy with THE Propeller chip.
I am still here, not there... so does that answer the question.
As Peter said, the prop is a joy to use, both software and hardware wise. To save complexity, just add more standalone props and link them together. Take a look at the TriBladeProp. Also the standalone props designed just for a single (but complex) job - the VT100 Terminal on a pcb for use with other micros, this website (becuase I am not sure how much I can reveal) www.propgfx.co.uk and others.
My TriBladeProp uses one prop for I/O expansion - it is a simple and actually cost effective way for a multipurpose, fully programmable, I/O peripheral. Another prop is used for the VT100 (or whatever) standalone peripheral in a single chip. I have provided the prop with extra SRAM should we want better graphics. Once again, this is a cost effective way and reduces the complexity for the programmer. The third prop has SRAMs and microSD for all sorts of powerful code. Once again this is a cost effective way to do this and gives us 8 cogs to do whatever we want. Things like PropDOS for a prop SBC (Single Board Computer), Z80 + CP/M (heaters iCog), and other emulations come to mind.
I have used co-operating microprocessor techniques since the early 80's. As I said above, it is a cost effective way to add functions to the processor and saves the complexity for the programmer. And we are using the same micro so only 1 set of code to learn.
The prop is a very powerful chip and with 8 cogs (real 32 bit processors) and the simple interface, we have a brilliant chip. Add a few together and you have a very simple architecture, easy to maintain, and easy to code.
Hope this says why I am still here having a ball
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Links to other interesting threads:
Xmos looks like an interesting chip, but I am staying away from it and concentrating on the Propeller. Why?
- Xmos "on-line" compiler for the inexpensive dev kit. Sending all my source to their web site... I think not.
- $999 "more complete" dev kit / compiler. I think not.
- availability of 40 pin dip propeller. Add a handful of parts, and you have a new board.
If Xmos was available as a dip part, with a free/cheap (off-line) dev kit, I'd look at it, but so far, I *REALLY* like the Prop.
ianw said...
Not to reopen an old can of worms, but how are propeller enthusiasts getting along with this chip now that they have given it some of their time?
I saw this post on the XMOS forums (back in October) regarding several propeller enthusiasts active in these forums, apparently crossing over to XMOS: www.xlinkers.org/forum/viewtopic.php?f=3&t=93&st=0&sk=t&sd=a
Although Leon and Ale have developed some code for the XMOS and Leon is probably more active in the XMOS forums now than the propeller forums, several of the other folks (heater coley, timothy, cluso) mentioned in the thread above seem to have not gone much further with the XMOS. Was productivity trickier than the Propeller?
Their is a nice set of code examples accumulating in their obex: www.xlinkers.org/projlist
But it is still no where near the level that is in the Propeller Obex
I've been working with two other regular posters to these forums on a multi rotor craft and one recently decided the XMOS was a better solution for his efforts than the Propeller. We debated crossing over to the XMOS with him, but will stick with the propeller temporarily. I think all three of us are curious how the pioneers listed above have found the development experience with the XMOS chips relative to the Propeller though.
In a previous post by Leon he indicated there will a single cog version available in volume at $1. This does seem very interesting. For example I may want to use Propeller code for a task, but only need 1-3 cogs and can't justify a $15 chip, especially if considering volume. But I could not find this statement on their forums, so was not sure how that price was derived - apparently it will be released later this year.
Looking at the underside of the evaluation pcb posted by timothy, it seems there are many more component dependancies for an XMOS setup than Propeller. The other concern is that this requires a 6 to 8 layer pcb, and that takes me outside the realm of free versions of Eagle or Target3001. www.flickr.com/photos/brilldea/2963394963/sizes/l/
You can get a more complete dev kit too, the XDK, but are they serious? $999 - that's not hobbyist friendly
I'd like to ask how the Propeller II will match against this, since I might be less tempted to stray, but I know that's taboo.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ www.mikronauts.com - a new blog about microcontrollers
For those of us who are citing Propeller prices, I should let you know that they're going to get really better as soon as Lauren lets loose (Qty 1 of any package: $7.99). Right now we're in the mandatory "distributor notification period" and once that's done you'll know about it officially!
As for the XMOS comparison, there's so much more than specifications. If all that people cared about were side-by-side specifications and they migrated to the "most capable" device, the BASIC Stamp would have been dead a long time ago but it's still finding new customers.
Ken Gracey (Parallax) said...
For those of us who are citing Propeller prices, I should let you know that they're going to get really better as soon as Lauren lets loose (Qty 1 of any package: $7.99). Right now we're in the mandatory "distributor notification period" and once that's done you'll know about it officially!
Now we are cooking with gas! With the way the exchange rate has headed for us antipodeans I was starting to have to balance the ease of use against the price. This sweetens the deal considerably. Now I don't suppose this might follow on to a price drop of the Proto Boards by any chance ?? [noparse];)[/noparse]
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cardinal Fang! Fetch the comfy chair.
Cluso, You are a genius! You wrote "heaters iCog". I was trying to think up a name for a cut down Intel 8080/5 only version of the ZiCog emulator. "iCog" is obviously it. Or was that a typo[noparse];)[/noparse]
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
I did have a brief flirtation with XMOS but I found it's IDE clunky, it's nowhere near as user friendly as the Propeller IDE.
I looked at it for an XMOS derivative of PropGFX but it's video generating capabilities are not as efficient in terms of resource allocation as the Propeller.
I was completely drawn in by the 400 mips with 8 threads on 4 tiles which sounded like a dream but the reality was somewhat different.
I was also shocked by the heat generated from the chip itself, seriously those things can burn!
There is no doubt it is a very capable piece of hardware but it's not for me.
The truth is I just missed my first love the Propeller too much, it's soooo easy to work with.
I'll just be patient and wait for Prop II (hint, hint, Chip!)
Coley
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ PropGFX - The home of the Hybrid Development System and PropGFX Lite
ianw
The single core XMOS chip will be available in a couple of months (QFN package), and a double sided PCB will be suitable. The eight hardware threads will be ample for many applications, giving a total of 400 MIPS. Having 64 I/Os will make it suitable for lots of interesting applications. The $1 price is for large quantities (like 250,000). A single chip will probably be about the same price as the Propeller.
It is feasible to put the BGA XMOS chips on a four-layer PCB, which isn't too bad.
The hardware works as advertised. The free tools include a debugger and simulator as well as XC and C compilers, and are pretty good. I am beta-testing the new timing analyser, which helps in the development of true deterministic code.
As with the Propeller, functions usually implemented in hardware are implemented in software. High-speed USB and Ethernet as well as the usual SPI, I2C and async comms are available as free downloads, as well as lots of other stuff.
Ale is the most productive Propeller user working with the XMOS chip, having implemented some nice VGA applications. The main limitation is the 64k of memory available on each core, which restricts the resolution. It isn't a limitation with most applications.
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
You got it in one there Gadgetman, plus this forum, is full of amazingly talented, and helpful people, which is probably why Leon keeps coming back here [noparse]:D[/noparse] lol
You just can't beat the Propeller for soooooooooooo many reasons, yes, it may not be the fastest chip in town, or have the most memory, but it has SOOOOOOOOOOOO much more to offer than any other micro I've worked on, heck, it's even more enjoyable than any games console I've worked on, and believe me, I've worked on all the mainstream consoles, and enjoyed them all, but nowhere near as much as I've enjoyed working on the propeller.
Thanks guys, and not just the Parallax people, but all on here too. [noparse]:D[/noparse]
@William. No, there is no relation and these are two mutually exclusive, independent events. As you have asked many times, and we have answered: there is no release date for Prop II. It's not even laid out yet. There is no release date. We don't know the cost or price of it. We will tell our customers when it is available.
I'm not just a Propeller "yes man", I do in fact read everything I can get my hands on regarding micro controllers. Parallax has set the bar very high.
The Propeller is the best bang for the buck and easiest to use for the level of functionality it provides.
The only processor I see that might surpass it will be the Propeller II.
I can think of a number of (personally) negatives for the Propeller and tools but if anyone asked me which chips I'd prefer to be working with day-in, day-out it would still be at the top of my list.
It's the best micro I've found for Rapid Application Development and prototyping, and I love the clean, uncomplicated architecture. If nothing else, the less there is, the less you have to learn. The great thing for me is that you can go from throwing something quick and dirty together in Spin to fine-tuning it to high-speed PASM or LMM.
The prop is great for loads of reasons, simplicity is one of them, hardware and soft-wise. OK, it is multicore, but if that is a problem for you (not yet multi-core aware brains without latest updates [noparse]:D[/noparse]), just think of them as interrupt driven and voila, it is as if it were [noparse]:D[/noparse].
The XMOS device is something else, tools are gcc/gdb based. That means quite a bit if you are custom to gnu tools (like me). But it has some niceties: a cycle exact simulator! (slow but useful to catch exceptions), and a time analysis tool as Leon said, and the tools work not only on winblows but also on MacOS X and Linux natively. Well, now that we have BST (thanks again Brad!), that is a no issue to me.
The online IDE... I never used it that much, I prefer local tools.
Hardware wise, well you have to do almost everything with software, but the ports are a tad more flexible than the Prop's. It is a different beast. Do not compare them, because they are for different purposes.
I just received my first Propeller (Demo board), and I'm currently comparing it to other options for low-power 4-channel audio processing for acoustic localisation (I basically need to do a convolution on 4 simultaneous 24-bit audio streams at 96kHz), at lowest possible power consumption and lowest possible complexity and size.
Reading through this discussion was very interesting, as it scratches a few of the things that I'm concerned about.
There's a few things that I found difficult to compare, that came up a couple of times:
MIPS performance: comparing the "speed" is quite difficult, and obviously the XMOS has far more MIPS, at higher power consumption, higher price and higher complexity. So does the ARM7 running at 400Mhz, or the Analog Devices BlackFin Dual Core running at 800Mhz. They are different CPUs for different purposes - nobody would use a BlackFin to build a weather station.
In terms of multi-CPU I'd like to mention the PowerPC MPC series, i.e. MPC5554 and similar processors - it's only single core, but has multiple TPUs (time processor units) which are freely microprogrammable parallel processors, that allow cycle-accurate real time response, and can be linked together. In terms of "programmable peripherals" that gets quite close, but the chips are far more expensive, and board design is more complex. But again, it's a far bigger, more powerful processor for completely different things. But then, it does have support for the Ada language [noparse]:)[/noparse]
In terms of power consumption I find the Atmel AVR32 comparable, which is also comparable in price, and also has roughly the same speed in MIPS as it seems. It's only single core, so it doesn't answer the original question, but it has heaps of peripherals that can lessen the computational load. See next point.
Hardware peripherals vs. Cogs. I have to apologize if I get some things wrong and please correct me - I obviously don't have much experience with the propeller yet, but have worked on a range of other platforms (mostly AVR, PowerPC and ARM).
As a Propeller newbie, I tried to find out what can be achieved. I know that an 8-bit Atmel can run SPI at 10Mhz using its dedicated hardware, or up to clock rate on the AVR32s. - I checked OBEX and found the SPI drivers, but none specified what speed can be supported. Same goes for other interfaces (i2c, UART, i2s, etc).
Another example is hardware timers. By cleverly linking multiple HW timers, interrupts and output compare units, one can "easily" generate clock cycle accurate signals with zero jitter, while the cpu can go to sleep. Of course I can do that on a cog, but wouldn't the duration of instructions (4 cycles?) limit my timing resolution and accuracy?
My point is, the configurability of the Propeller is great, but needs MIPS - so it makes comparisons harder, and it's not quite clear without trying it out how much performance one can expect. Maybe someone can enlighten me [noparse]:)[/noparse]
Interrupts and determinism:
I think it's important to distinguish between external and internal interrupts here. It is true that interrupts mostly ruin determinism to some extent (mostly introducing timing jitter, so in the end reducing timing resolution, but sometimes can really upset the system as well). However, it is possible to write code that uses interrupts and yet is 100% deterministic, i.e. if the interrupt is caused by an internal timer that is set appropriately, and the main loop is empty (idle). It does need some analysis to make sure that the interrupts don't interrupt each other. One application I used the timer interrupt to program the HW timer output compare for the next cycle, and was triggered after the next output event. That way the output edge was always 100% accurate and aligned with the system clock. Jitter was not visible, and probably in the low nanosec range.
Having said all that, i think the Propeller is a fantastically exciting processor, and I can see me using it as a much better, faster and more flexible alternative to the ubiquitous ATMegas in my designs.
I think Parallax could make the propeller more appealing to professionals by demonstrating the performance more clearly and showing what can be done at what accuracy and speed and what can't. Especially the performance of emulated peripherals is crucial to me if I want to connect high-speed ADC/DACs or other peripherals. While I can easily check the data sheet of i.e. the AVR32 UC3 or AP7000 to read about SPI clock rates, I2S support etc, I found it difficult to get that information for the propeller.
Some comments:
"Hardware peripherals vs. Cogs"
By cleverly linking cog timers and cogs, syncing multiple cogs, and/or possibly communicating via otherwise unused I/O pins, one can "easily" do all sorts of very complex signal generation to a resolution of one clock cycle on the Propeller. Yes, the basic instruction time is 4 clock cycles (50ns), but, depending on the particular application, you can get quite clever about doing things faster using more resources. The use of multiple cogs (6) to get 1600 x 1200 pixel VGA output is one example.
"Interrupts and determinism"
The WAITCNT instruction does allow a cog to synchronize itself to a specific clock cycle and the WAITPNE/WAITPEQ instruction allows a cog to synchronize itself to a specific I/O state, both with a granularity of a single clock cycle and complete determinism.
You're correct about the value of lots of application notes and similar documentation. As is usual for small companies, Parallax runs a "tight ship" and people tend to wear many hats and stay very busy. There's simply not enough resources (very talented and experienced people) available for the sort of documentation you're describing to be done in anything other than years of one little side project at a time.
One difficulty about using software to implement functional blocks over functional blocks in hardware is that things are so flexible. Usually there's a constant tradeoff of speed vs. functionality or memory space that doesn't exist with hardware functional blocks. For example, there are now 3 or 4 ways at least of doing buffered asynchronous serial I/O with tradeoffs of code size, maximum Baud, other features like handshaking, etc. There are several ways of doing I2C, several ways of doing SPI, etc. Really high performance applications tend to integrate the I/O functionality into the rest of the code rather than use more general purpose library routines, so general statements in a datasheet may not apply.
Welcome to the prop community. I think you'll find all of what Mike Green said to be true. And if you spend any amount of time on the forums, you'll quickly learn that he really knows his stuff. He was favorably compared to Santa Claus last Christmas because of how much he gives to everyone here. I've often been on the recieving end of Mike's advice.
The project you briefly describe will be challenging, but I believe it can be done in the Prop (with Assembly). I have a device based on the Prop which reads 4 channels of 24-bit data from audio ADCs. These data are read and averaged then dumped to HUB RAM for other cogs to work on. The ACQ and averaging takes 2 cogs for all 4 channels. @96Khz we were just able to get our 24-bits out of the 32-bit output word (we're using the AD1871. What ADC are you using?). The lack of a PASM multiply will makes this a bit more complicated, but there are easy to find mul routines in PASM which will do this. Depending on how many other jobs this Prop has to handle (driving display, data storage, etc) you may have enough resources to do real-time continuous analysis.
When you get your project going, or hit a snag, be sure to post on a new thread so we can see how this project comes along.
I don't think that convolution for acoustic localisation is feasible on 24-bit audio streams at 96 kHz, on the Propeller. There simply isn't enough memory.
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Thanks for the quick and detailed responses.
I haven't fully decided on the ADC yet - probably either TI's ADS1274 (has SPI data interface, industrial grade ADC with very good SNR and low drift) or one of the audio CODECs (not 100% sure if I need a dedicated DAC as well).
It's good to know that it seems to be possible to get the data rate into the chip. I assume you were using two chips - was that to be able to use two cogs independently to grab the data? would it be possible to read a single SPI stream with all 4 channels, or is that too much?
In terms of memory, 32 +8x 2kb Kb isn't that much (definitely looking forward to a Propeller 2 with more onboard RAM [noparse]:)[/noparse] ), but I think it'll do. I expect maybe 10 samples maximum search width (pretty short baseline) and a window size of 256 samples at most. As everything is running in real time there is no need to keep any more data than the last windowsize plus search size (unless I want to use some pre-calc optimisations to speed it up).
The main concern I have though is about the lack of hardware muls. My other choice at the moment is the AVR32 (AT32UC3A1256) which has 64 Kb RAM, 1-cycle instructions, muls, and of course all the peripherals. It's only spec'ed at 90 MIPS (@66Mhz), so it's a bit slower, but given the multiplication issue it might actually be faster for my application...
Maybe I just have to make two test boards, one with Propeller and one with AVR32 and see which one wins [noparse]:)[/noparse]
Of course a slightly bigger processor might be better suited, but I'm trying to avoid the complexity of external RAM, BGA packages, complicated boot setups, memory map configurations etc. (and I don't want an OS running as I need fairly high sync accuracy).
I have been a Parallax family member since way before the forum. I like Parallax, and their products.
I do not knock the Xmos products, but I know little about them. I don't think I will either with the price tag they have hung on the header of that board.
I have looked around, and considered many other development systems, but I always return to Parallax. Some of their items are a little pricey, but pound for pound, dollar for dollar, you will not find better customer support, nor an easier product to start with.
I'm sure there are many other systems out there that can do this or that, but the product is not all I look at.
I'm late into this discussion, but I was around way before the Propeller, and will be long after Propeller II.
Comments
I saw this post on the XMOS forums (back in October) regarding several propeller enthusiasts active in these forums, apparently crossing over to XMOS:
www.xlinkers.org/forum/viewtopic.php?f=3&t=93&st=0&sk=t&sd=a
Although Leon and Ale have developed some code for the XMOS and Leon is probably more active in the XMOS forums now than the propeller forums, several of the other folks (heater coley, timothy, cluso) mentioned in the thread above seem to have not gone much further with the XMOS. Was productivity trickier than the Propeller?
Their is a nice set of code examples accumulating in their obex: www.xlinkers.org/projlist
But it is still no where near the level that is in the Propeller Obex
I've been working with two other regular posters to these forums on a multi rotor craft and one recently decided the XMOS was a better solution for his efforts than the Propeller. We debated crossing over to the XMOS with him, but will stick with the propeller temporarily. I think all three of us are curious how the pioneers listed above have found the development experience with the XMOS chips relative to the Propeller though.
In a previous post by Leon he indicated there will a single cog version available in volume at $1. This does seem very interesting. For example I may want to use Propeller code for a task, but only need 1-3 cogs and can't justify a $15 chip, especially if considering volume. But I could not find this statement on their forums, so was not sure how that price was derived - apparently it will be released later this year.
Looking at the underside of the evaluation pcb posted by timothy, it seems there are many more component dependancies for an XMOS setup than Propeller. The other concern is that this requires a 6 to 8 layer pcb, and that takes me outside the realm of free versions of Eagle or Target3001.
www.flickr.com/photos/brilldea/2963394963/sizes/l/
You can get a more complete dev kit too, the XDK, but are they serious? $999 - that's not hobbyist friendly
I'd like to ask how the Propeller II will match against this, since I might be less tempted to stray, but I know that's taboo.
- Parallax is recognized globally.
- Many years in the business.
- Distributors throughout the world.
- Propeller is easy to use (Software and Hardware).
- Only 13 dollars (if cost more I would not buy it, and I expect that its price descend, since is a mature product).
- Propeller Protoboard at 19 dollars
(Now 32.00)
- The support of all the community.
- The version for Linux/MacOS (Thanks Brad).
But certainly XMOS has something very powerful, and we waiting since years ago the Propeller II, as was promised.Good luck to the XMOS boys and I may hop on that wagon if it's going my way but my love is the Prop chip. It is both a "hobbyist" chip and also a very serious (but fun) commercial chip. Viva la Prop!
Prop II will be here before we know it but a watched kettle never boils, so keep busy with THE Propeller chip.
*Peter*
As Peter said, the prop is a joy to use, both software and hardware wise. To save complexity, just add more standalone props and link them together. Take a look at the TriBladeProp. Also the standalone props designed just for a single (but complex) job - the VT100 Terminal on a pcb for use with other micros, this website (becuase I am not sure how much I can reveal) www.propgfx.co.uk and others.
My TriBladeProp uses one prop for I/O expansion - it is a simple and actually cost effective way for a multipurpose, fully programmable, I/O peripheral. Another prop is used for the VT100 (or whatever) standalone peripheral in a single chip. I have provided the prop with extra SRAM should we want better graphics. Once again, this is a cost effective way and reduces the complexity for the programmer. The third prop has SRAMs and microSD for all sorts of powerful code. Once again this is a cost effective way to do this and gives us 8 cogs to do whatever we want. Things like PropDOS for a prop SBC (Single Board Computer), Z80 + CP/M (heaters iCog), and other emulations come to mind.
I have used co-operating microprocessor techniques since the early 80's. As I said above, it is a cost effective way to add functions to the processor and saves the complexity for the programmer. And we are using the same micro so only 1 set of code to learn.
The prop is a very powerful chip and with 8 cogs (real 32 bit processors) and the simple interface, we have a brilliant chip. Add a few together and you have a very simple architecture, easy to maintain, and easy to code.
Hope this says why I am still here having a ball
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Links to other interesting threads:
· Home of the MultiBladeProps: TriBladeProp, SixBladeProp, website (Multiple propeller pcbs)
· Prop Tools under Development or Completed (Index)
· Emulators (Micros eg Altair, and Terminals eg VT100) - index
· Search the Propeller forums (via Google)
My cruising website is: ·www.bluemagic.biz·· MultiBladeProp is: www.bluemagic.biz/cluso.htm
- Xmos "on-line" compiler for the inexpensive dev kit. Sending all my source to their web site... I think not.
- $999 "more complete" dev kit / compiler. I think not.
- availability of 40 pin dip propeller. Add a handful of parts, and you have a new board.
If Xmos was available as a dip part, with a free/cheap (off-line) dev kit, I'd look at it, but so far, I *REALLY* like the Prop.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com - a new blog about microcontrollers
As for the XMOS comparison, there's so much more than specifications. If all that people cared about were side-by-side specifications and they migrated to the "most capable" device, the BASIC Stamp would have been dead a long time ago but it's still finding new customers.
Ken Gracey
Parallax, Inc.
Now we are cooking with gas! With the way the exchange rate has headed for us antipodeans I was starting to have to balance the ease of use against the price. This sweetens the deal considerably. Now I don't suppose this might follow on to a price drop of the Proto Boards by any chance ?? [noparse];)[/noparse]
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cardinal Fang! Fetch the comfy chair.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
For me, the past is not over yet.
Proto boards are free for you because of your Mac/Linux efforts, so you can consider that a "price drop". Just tell me how many you need!
Ken Gracey
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.fd.com.my
www.mercedes.com.my
I looked at it for an XMOS derivative of PropGFX but it's video generating capabilities are not as efficient in terms of resource allocation as the Propeller.
I was completely drawn in by the 400 mips with 8 threads on 4 tiles which sounded like a dream but the reality was somewhat different.
I was also shocked by the heat generated from the chip itself, seriously those things can burn!
There is no doubt it is a very capable piece of hardware but it's not for me.
The truth is I just missed my first love the Propeller too much, it's soooo easy to work with.
I'll just be patient and wait for Prop II (hint, hint, Chip!)
Coley
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
PropGFX - The home of the Hybrid Development System and PropGFX Lite
The single core XMOS chip will be available in a couple of months (QFN package), and a double sided PCB will be suitable. The eight hardware threads will be ample for many applications, giving a total of 400 MIPS. Having 64 I/Os will make it suitable for lots of interesting applications. The $1 price is for large quantities (like 250,000). A single chip will probably be about the same price as the Propeller.
It is feasible to put the BGA XMOS chips on a four-layer PCB, which isn't too bad.
The hardware works as advertised. The free tools include a debugger and simulator as well as XC and C compilers, and are pretty good. I am beta-testing the new timing analyser, which helps in the development of true deterministic code.
As with the Propeller, functions usually implemented in hardware are implemented in software. High-speed USB and Ethernet as well as the usual SPI, I2C and async comms are available as free downloads, as well as lots of other stuff.
Ale is the most productive Propeller user working with the XMOS chip, having implemented some nice VGA applications. The main limitation is the 64k of memory available on each core, which restricts the resolution. It isn't a limitation with most applications.
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Post Edited (Leon) : 3/12/2009 10:22:17 AM GMT
Makes using props as peripherals to props even better!!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Links to other interesting threads:
· Home of the MultiBladeProps: TriBladeProp, SixBladeProp, website (Multiple propeller pcbs)
· Prop Tools under Development or Completed (Index)
· Emulators (Micros eg Altair, and Terminals eg VT100) - index
· Search the Propeller forums (via Google)
My cruising website is: ·www.bluemagic.biz·· MultiBladeProp is: www.bluemagic.biz/cluso.htm
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Don't visit my new website...
You just can't beat the Propeller for soooooooooooo many reasons, yes, it may not be the fastest chip in town, or have the most memory, but it has SOOOOOOOOOOOO much more to offer than any other micro I've worked on, heck, it's even more enjoyable than any games console I've worked on, and believe me, I've worked on all the mainstream consoles, and enjoyed them all, but nowhere near as much as I've enjoyed working on the propeller.
Thanks guys, and not just the Parallax people, but all on here too. [noparse]:D[/noparse]
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
http://www.propgfx.co.uk/forum/·home of the PropGFX Lite
·
Ken Gracey
Parallax, Inc.
The Propeller is the best bang for the buck and easiest to use for the level of functionality it provides.
The only processor I see that might surpass it will be the Propeller II.
OBC
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
New to the Propeller?
Check out: Protoboard Introduction , Propeller Cookbook 1.4 & Software Index
Updates to the Cookbook are now posted to: Propeller.warrantyvoid.us
Got an SD card connected? - PropDOS
It's the best micro I've found for Rapid Application Development and prototyping, and I love the clean, uncomplicated architecture. If nothing else, the less there is, the less you have to learn. The great thing for me is that you can go from throwing something quick and dirty together in Spin to fine-tuning it to high-speed PASM or LMM.
@BradC, sounds like you've got Protoboards for your next projects.. [noparse]:)[/noparse]
OBC
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
New to the Propeller?
Check out: Protoboard Introduction , Propeller Cookbook 1.4 & Software Index
Updates to the Cookbook are now posted to: Propeller.warrantyvoid.us
Got an SD card connected? - PropDOS
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Links to other interesting threads:
· Home of the MultiBladeProps: TriBladeProp, SixBladeProp, website (Multiple propeller pcbs)
· Prop Tools under Development or Completed (Index)
· Emulators (Micros eg Altair, and Terminals eg VT100) - index
· Search the Propeller forums (via Google)
My cruising website is: ·www.bluemagic.biz·· MultiBladeProp is: www.bluemagic.biz/cluso.htm
The XMOS device is something else, tools are gcc/gdb based. That means quite a bit if you are custom to gnu tools (like me). But it has some niceties: a cycle exact simulator! (slow but useful to catch exceptions), and a time analysis tool as Leon said, and the tools work not only on winblows but also on MacOS X and Linux natively. Well, now that we have BST (thanks again Brad!), that is a no issue to me.
The online IDE... I never used it that much, I prefer local tools.
Hardware wise, well you have to do almost everything with software, but the ports are a tad more flexible than the Prop's. It is a different beast. Do not compare them, because they are for different purposes.
Did I said something new ?
Reading through this discussion was very interesting, as it scratches a few of the things that I'm concerned about.
There's a few things that I found difficult to compare, that came up a couple of times:
In terms of multi-CPU I'd like to mention the PowerPC MPC series, i.e. MPC5554 and similar processors - it's only single core, but has multiple TPUs (time processor units) which are freely microprogrammable parallel processors, that allow cycle-accurate real time response, and can be linked together. In terms of "programmable peripherals" that gets quite close, but the chips are far more expensive, and board design is more complex. But again, it's a far bigger, more powerful processor for completely different things. But then, it does have support for the Ada language [noparse]:)[/noparse]
In terms of power consumption I find the Atmel AVR32 comparable, which is also comparable in price, and also has roughly the same speed in MIPS as it seems. It's only single core, so it doesn't answer the original question, but it has heaps of peripherals that can lessen the computational load. See next point.
As a Propeller newbie, I tried to find out what can be achieved. I know that an 8-bit Atmel can run SPI at 10Mhz using its dedicated hardware, or up to clock rate on the AVR32s. - I checked OBEX and found the SPI drivers, but none specified what speed can be supported. Same goes for other interfaces (i2c, UART, i2s, etc).
Another example is hardware timers. By cleverly linking multiple HW timers, interrupts and output compare units, one can "easily" generate clock cycle accurate signals with zero jitter, while the cpu can go to sleep. Of course I can do that on a cog, but wouldn't the duration of instructions (4 cycles?) limit my timing resolution and accuracy?
My point is, the configurability of the Propeller is great, but needs MIPS - so it makes comparisons harder, and it's not quite clear without trying it out how much performance one can expect. Maybe someone can enlighten me [noparse]:)[/noparse]
I think it's important to distinguish between external and internal interrupts here. It is true that interrupts mostly ruin determinism to some extent (mostly introducing timing jitter, so in the end reducing timing resolution, but sometimes can really upset the system as well). However, it is possible to write code that uses interrupts and yet is 100% deterministic, i.e. if the interrupt is caused by an internal timer that is set appropriately, and the main loop is empty (idle). It does need some analysis to make sure that the interrupts don't interrupt each other. One application I used the timer interrupt to program the HW timer output compare for the next cycle, and was triggered after the next output event. That way the output edge was always 100% accurate and aligned with the system clock. Jitter was not visible, and probably in the low nanosec range.
Having said all that, i think the Propeller is a fantastically exciting processor, and I can see me using it as a much better, faster and more flexible alternative to the ubiquitous ATMegas in my designs.
I think Parallax could make the propeller more appealing to professionals by demonstrating the performance more clearly and showing what can be done at what accuracy and speed and what can't. Especially the performance of emulated peripherals is crucial to me if I want to connect high-speed ADC/DACs or other peripherals. While I can easily check the data sheet of i.e. the AVR32 UC3 or AP7000 to read about SPI clock rates, I2S support etc, I found it difficult to get that information for the propeller.
Post Edited (sfx) : 4/26/2009 9:27:52 AM GMT
"Hardware peripherals vs. Cogs"
By cleverly linking cog timers and cogs, syncing multiple cogs, and/or possibly communicating via otherwise unused I/O pins, one can "easily" do all sorts of very complex signal generation to a resolution of one clock cycle on the Propeller. Yes, the basic instruction time is 4 clock cycles (50ns), but, depending on the particular application, you can get quite clever about doing things faster using more resources. The use of multiple cogs (6) to get 1600 x 1200 pixel VGA output is one example.
"Interrupts and determinism"
The WAITCNT instruction does allow a cog to synchronize itself to a specific clock cycle and the WAITPNE/WAITPEQ instruction allows a cog to synchronize itself to a specific I/O state, both with a granularity of a single clock cycle and complete determinism.
You're correct about the value of lots of application notes and similar documentation. As is usual for small companies, Parallax runs a "tight ship" and people tend to wear many hats and stay very busy. There's simply not enough resources (very talented and experienced people) available for the sort of documentation you're describing to be done in anything other than years of one little side project at a time.
One difficulty about using software to implement functional blocks over functional blocks in hardware is that things are so flexible. Usually there's a constant tradeoff of speed vs. functionality or memory space that doesn't exist with hardware functional blocks. For example, there are now 3 or 4 ways at least of doing buffered asynchronous serial I/O with tradeoffs of code size, maximum Baud, other features like handshaking, etc. There are several ways of doing I2C, several ways of doing SPI, etc. Really high performance applications tend to integrate the I/O functionality into the rest of the code rather than use more general purpose library routines, so general statements in a datasheet may not apply.
Welcome to the prop community. I think you'll find all of what Mike Green said to be true. And if you spend any amount of time on the forums, you'll quickly learn that he really knows his stuff. He was favorably compared to Santa Claus last Christmas because of how much he gives to everyone here. I've often been on the recieving end of Mike's advice.
The project you briefly describe will be challenging, but I believe it can be done in the Prop (with Assembly). I have a device based on the Prop which reads 4 channels of 24-bit data from audio ADCs. These data are read and averaged then dumped to HUB RAM for other cogs to work on. The ACQ and averaging takes 2 cogs for all 4 channels. @96Khz we were just able to get our 24-bits out of the 32-bit output word (we're using the AD1871. What ADC are you using?). The lack of a PASM multiply will makes this a bit more complicated, but there are easy to find mul routines in PASM which will do this. Depending on how many other jobs this Prop has to handle (driving display, data storage, etc) you may have enough resources to do real-time continuous analysis.
When you get your project going, or hit a snag, be sure to post on a new thread so we can see how this project comes along.
Best of luck,
Peter
Leon
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle
Post Edited (Leon) : 4/26/2009 6:30:55 PM GMT
I haven't fully decided on the ADC yet - probably either TI's ADS1274 (has SPI data interface, industrial grade ADC with very good SNR and low drift) or one of the audio CODECs (not 100% sure if I need a dedicated DAC as well).
It's good to know that it seems to be possible to get the data rate into the chip. I assume you were using two chips - was that to be able to use two cogs independently to grab the data? would it be possible to read a single SPI stream with all 4 channels, or is that too much?
In terms of memory, 32 +8x 2kb Kb isn't that much (definitely looking forward to a Propeller 2 with more onboard RAM [noparse]:)[/noparse] ), but I think it'll do. I expect maybe 10 samples maximum search width (pretty short baseline) and a window size of 256 samples at most. As everything is running in real time there is no need to keep any more data than the last windowsize plus search size (unless I want to use some pre-calc optimisations to speed it up).
The main concern I have though is about the lack of hardware muls. My other choice at the moment is the AVR32 (AT32UC3A1256) which has 64 Kb RAM, 1-cycle instructions, muls, and of course all the peripherals. It's only spec'ed at 90 MIPS (@66Mhz), so it's a bit slower, but given the multiplication issue it might actually be faster for my application...
Maybe I just have to make two test boards, one with Propeller and one with AVR32 and see which one wins [noparse]:)[/noparse]
Of course a slightly bigger processor might be better suited, but I'm trying to avoid the complexity of external RAM, BGA packages, complicated boot setups, memory map configurations etc. (and I don't want an OS running as I need fairly high sync accuracy).
I do not knock the Xmos products, but I know little about them. I don't think I will either with the price tag they have hung on the header of that board.
I have looked around, and considered many other development systems, but I always return to Parallax. Some of their items are a little pricey, but pound for pound, dollar for dollar, you will not find better customer support, nor an easier product to start with.
I'm sure there are many other systems out there that can do this or that, but the product is not all I look at.
I'm late into this discussion, but I was around way before the Propeller, and will be long after Propeller II.
James L
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
James L
Partner/Designer
Lil Brother SMT Assembly Services