I believe that there are definite advantages to other designs, but definite drawbacks also. I think it is wise to take the best options from each.
The SeaForth chips, for instance, have an order of magnitude(or 2) faster core to core communications, but 18 bits is quite odd, and the local memory to each core is really small, the Forth community is quite small, support is quite limited, and the dev kit is expensive, and the packaging is not friendly. But, if you take a look at the bandwidth you can pass through this tiny chip, and it's processing power, it is quite amazing. The links to other cores is similar to the transputer, which was used as a communications grid to tie large arrays of faster processors together in the day. XMOS uses the same kind of technology in the current chips, and we here about them every time Leon says anything on this forum.
As far as the OMAP3530, I thought it was you(or maybe it was BradC) that pointed out the beagleboard, which is based on it. There are more videos of it on youtube than on the propeller. OMAP4 should be spectacular for such a tiny low power thing. It is closer to the opposite side of the complexity scale, but not as far as the x86 and competing processors.
To add up down left and right links would not add that much complexity, but it would add bandwidth. It would take away 4 longs from the 2k of cog ram. Remember turbulance, the author (I forget) used 8 or so IO pins as a bus between the cogs. This is only a niche case because we currently have to sacrifice precious i/o pins to do it. Even if portb worked on the current prop without external pins, you would be seeing lots of people using as a databus. Sadly, it is one of the shortcommings/wasted silicon, of the current prop.
I believe that the main strengths of the propeller is the lack of interrupts, simple design, and this forum. The SeaForth has no interrupts, has a simple design(even simpler), but does not have this forum.
I would agree that the hub provides a simple interface that makes putting together objects easy. It is also pretty amazing what has been done with it.
I can't wait til propII. Then things will get VERY amazing.
The Z80 is kinda like the Propeller Processor in away. It was way ahead of its time and got a whole generation
of young engineers interested in computer systems.
I will never forget trying to learn basic on my Sinclair ZX80, what a pain. [noparse]:)[/noparse]...
As for me, I need to get bursts of computation very quickly, mostly to conserve battery power, so Spin is out of the question.. just too slow. I have now managed to write several small kernels (optimized for speed or size) to enable multiple threads per cog without significant interaction among the threads. Theoretically 128 threads in one cog... but that is not for a very realistic application. So to me, numers of cogs is now relatively insignificant, and would prefer to have some more memory per cog, but of course that screws up the addressing. In the extreme, how about a scaled down prop -based on prop1 low leakage (battery friendly) technology- with only two cogs thus a smaller less costly die, but with with a faster clock, and a pipeline for one instruction per clock.
Then, provide for at least one (two would be nicer) REAL indirect cog memory access, an OTP internal serial number, and we'd have a VERY good and improved replacement for the SX series. Add a general purpose clocked shift register with programmable feedback connections to generate 32 bit LFSR's or Manchester outputs for communication and I'd be satisfied for a VERY long time.
While we are dreaming... 4 cogs, 64KB+ hub ram, internally implemented Port B - in a DIP28 package, with 23 I/O's (Vcc, Vss, /RST, OSC1, OSC2 and P0-P22)
With single-cycle execution, it would be a perfect SX28 replacement!
And if it had at least 80KB hub ram... single chip Z80 system!
pjv said...
Hi All;
As for me, I need to get bursts of computation very quickly, mostly to conserve battery power, so Spin is out of the question.. just too slow. I have now managed to write several small kernels (optimized for speed or size) to enable multiple threads per cog without significant interaction among the threads. Theoretically 128 threads in one cog... but that is not for a very realistic application. So to me, numers of cogs is now relatively insignificant, and would prefer to have some more memory per cog, but of course that screws up the addressing. In the extreme, how about a scaled down prop -based on prop1 low leakage (battery friendly) technology- with only two cogs thus a smaller less costly die, but with with a faster clock, and a pipeline for one instruction per clock.
Then, provide for at least one (two would be nicer) REAL indirect cog memory access, an OTP internal serial number, and we'd have a VERY good and improved replacement for the SX series. Add a general purpose clocked shift register with programmable feedback connections to generate 32 bit LFSR's or Manchester outputs for communication and I'd be satisfied for a VERY long time.
microcontrolled said...
I notice that, now that the general PropII specs are out, people are already having PropIII dreams. Will it ever end?!?! NO other prop will be this low cost. This is the best Prop out there for high volume orders. I seriously think that until you get Win/Mac/Linux running on one all simultaniously with 50 USB ports 3TB memory and 400 GB RAM no one will be satisfyed. This cycle will continue until you can at least play HALO on your Prop with nothing but a screen and gamepad.
I remember saying something like this to Chip last year when he asked me how much memory we wanted. [noparse]:)[/noparse]
The week after PropII comes out, a thread for PropIII will be started by someone. -- There's no end. [noparse]:)[/noparse]
I'll be lucky to completely master the PropI by the time PropV makes it's debut!
(BTW, love the title graphic on your blog! Nice job.)
@OBC: Thanks! Did that on GIMP. I better update now that I know others are reading it... Oh, and you are wrong, OBC. There's a thread on Prop 3 already!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔ Computers are microcontrolled.
Comments
The SeaForth chips, for instance, have an order of magnitude(or 2) faster core to core communications, but 18 bits is quite odd, and the local memory to each core is really small, the Forth community is quite small, support is quite limited, and the dev kit is expensive, and the packaging is not friendly. But, if you take a look at the bandwidth you can pass through this tiny chip, and it's processing power, it is quite amazing. The links to other cores is similar to the transputer, which was used as a communications grid to tie large arrays of faster processors together in the day. XMOS uses the same kind of technology in the current chips, and we here about them every time Leon says anything on this forum.
As far as the OMAP3530, I thought it was you(or maybe it was BradC) that pointed out the beagleboard, which is based on it. There are more videos of it on youtube than on the propeller. OMAP4 should be spectacular for such a tiny low power thing. It is closer to the opposite side of the complexity scale, but not as far as the x86 and competing processors.
To add up down left and right links would not add that much complexity, but it would add bandwidth. It would take away 4 longs from the 2k of cog ram. Remember turbulance, the author (I forget) used 8 or so IO pins as a bus between the cogs. This is only a niche case because we currently have to sacrifice precious i/o pins to do it. Even if portb worked on the current prop without external pins, you would be seeing lots of people using as a databus. Sadly, it is one of the shortcommings/wasted silicon, of the current prop.
I believe that the main strengths of the propeller is the lack of interrupts, simple design, and this forum. The SeaForth has no interrupts, has a simple design(even simpler), but does not have this forum.
I would agree that the hub provides a simple interface that makes putting together objects easy. It is also pretty amazing what has been done with it.
I can't wait til propII. Then things will get VERY amazing.
Doug
of young engineers interested in computer systems.
I will never forget trying to learn basic on my Sinclair ZX80, what a pain. [noparse]:)[/noparse]...
Russ
As for me, I need to get bursts of computation very quickly, mostly to conserve battery power, so Spin is out of the question.. just too slow. I have now managed to write several small kernels (optimized for speed or size) to enable multiple threads per cog without significant interaction among the threads. Theoretically 128 threads in one cog... but that is not for a very realistic application. So to me, numers of cogs is now relatively insignificant, and would prefer to have some more memory per cog, but of course that screws up the addressing. In the extreme, how about a scaled down prop -based on prop1 low leakage (battery friendly) technology- with only two cogs thus a smaller less costly die, but with with a faster clock, and a pipeline for one instruction per clock.
Then, provide for at least one (two would be nicer) REAL indirect cog memory access, an OTP internal serial number, and we'd have a VERY good and improved replacement for the SX series. Add a general purpose clocked shift register with programmable feedback connections to generate 32 bit LFSR's or Manchester outputs for communication and I'd be satisfied for a VERY long time.
One can always dream.......
Cheers,
Peter (pjv)
With single-cycle execution, it would be a perfect SX28 replacement!
And if it had at least 80KB hub ram... single chip Z80 system!
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
Las - Large model assembler Largos - upcoming nano operating system
I remember saying something like this to Chip last year when he asked me how much memory we wanted. [noparse]:)[/noparse]
The week after PropII comes out, a thread for PropIII will be started by someone. -- There's no end. [noparse]:)[/noparse]
I'll be lucky to completely master the PropI by the time PropV makes it's debut!
(BTW, love the title graphic on your blog! Nice job.)
OBC
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
New to the Propeller?
Visit the: The Propeller Pages @ Warranty Void.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Computers are microcontrolled.
Robots are microcontrolled.
I am microcontrolled.
SX Spinning light display·
http://designedbymemicros.blogspot.com/
Sorry, I simply could not resist starting that thread!
However, once I started it... all my suggestions are perfectly serious, and what I would do.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
Las - Large model assembler Largos - upcoming nano operating system
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Computers are microcontrolled.
Robots are microcontrolled.
I am microcontrolled.
SX Spinning light display·
http://designedbymemicros.blogspot.com/
When is the Propeller II going to be available, we could use thousands of them right now.
Russ
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
Las - Large model assembler Largos - upcoming nano operating system