Good ideas, Everyone. That thing Phil showed is an apt analogy, but almost nobody knows the name of that thing. I think the planetary gear is a good visual analogy, with maybe a central gear that is the hub ram, surrounded by 16 small gears.
We had a meeting with OnSemi today at Parallax to update where we're all at, since we've embarked on this redesign. The path to completion is simpler than ever, but there's still a good amount of Verilog work to do. Hopefully, I'll have an FPGA image soon.
Yeah, I know, I've got the P1 image contest to kick off, and I'll do it this week I promise, but this idea of the gears could use some love while we all get through waiting for the FPGA.
We had a meeting with OnSemi today at Parallax to update where we're all at, since we've embarked on this redesign. The path to completion is simpler than ever, but there's still a good amount of Verilog work to do. Hopefully, I'll have an FPGA image soon.
Is the COG/HUB at a stage where OnSemi can run simulations on MHz and mA/MHz yet ?
I think that because we're used to thinking of memory as static, and pins connecting to the outside world around the periphery (and therefore cogs) as also static, its the connections that need to be very dynamic, while all the rest of the architecture remains static. The other thing is there needs to be multiple register/memory locations shown, which is relatively easy for an animated crosspoint diagram, but harder for other analogies (multiple movies inside the zoetrope, multiple dishes along each radial arm of the lazy susan. The train analogy is perhaps a little easier as "carriages" can be invoked
That means something like an animated crosspoint switch, or some approaching '70s "string art" might be the most applicable. We can put animations in pdfs now, right?
I did look for a 16 partner round dance but didn't succeed. The propeller do si do...
A Nibble sequential crosspoint is what the hardware does, which can be drawn as outers as COGs and inner as HUB memory, segmented into 16 Nibble fractions.
Minor tweaks to Chips drawing I'd do would be
* To merge/expand the yellow HUB memory (currently 16 isolated dots), to make it clearer it is not 16 smaller memories to the user.
* Present 2 (or more) copies to underscore the Rotating Physical model eg one @Cyc=0 and one @Cyc= +5
While looking at that drawing, I was reminded of a rotary-dial telephone. Do you think we could exploit that familiar concept? We'd just have to mention that each dial pulse has a 5ns cycle, instead of ~200ms. And, there are some other differences.
While looking at that drawing, I was reminded of a rotary-dial telephone. Do you think we could exploit that familiar concept? We'd just have to mention that each dial pulse has a 5ns cycle, instead of ~200ms. And, there are some other differences.
Hehe, Problems with that are
a) very few are going to know what a rotary-dial telephone looks like, or how it worked.
b) rotary-dial telephone have a single-pulse stream, the actual 16 way Mux connects all 16 COGS to one memory each, so it is important to preserve that in-parallel concept. The original image was very close.
Just show it in 2 positions to 'drive home' the (virtual) rotating cross point.
I was reminded of a rotary-dial telephone. Do you think we could exploit that familiar concept?
Familiar to whom? Old people like us?
The other day, a friend of mine was visiting with his 14-year-old daughter, I have a rotary dial phone in my living room -- mostly as a conversation piece -- but it's still hooked up. My friend's daughter asked, "Papa, how do you work one of these?" He had to demonstrate it for her. I asked her if she'd ever played a phonograph record. "Nope." The times, they are a-changin'.
OTOH, a dial telephone is a "little" more up-to-date than a zoetrope!
In CPU architecture we are familiar with pipelining so that the CPU doesn't have to wait for the next instruction although there are pipeline stalls, anyway these are concepts people in the know - know. We could refer to this hub memory as a Pipelined Multi-port RAM or PMR or just PRAM. Quite apt since it carries around Chip's baby
OTOH, a dial telephone is a "little" more up-to-date than a zoetrope!
-Phil
Prior to our dial telephone we had a telephone that responded to voice commands. You picked up the receiver and waited to be acknowledged by the operator, Blanche. You could give Blanche the number you wanted or you could just tell her who you wanted to speak with. The phone numbers were only three digits but 'Nan Hapgood' or 'Aunt Stella' were easier for a four year old to remember. Blanche had excellent voice recognition skills so she knew who you were when you spoke and who 'Nan Hapgood' or 'Aunt Stella' were based on recognizing your voice. Kind of a Star Trek thing before its time.
In the evenings Blanche might not be home so sometimes you would have to wait until morning to make a call. No big deal.
While looking at that drawing, I was reminded of a rotary-dial telephone. Do you think we could exploit that familiar concept? We'd just have to mention that each dial pulse has a 5ns cycle, instead of ~200ms. And, there are some other differences.
once there were daisy wheel printers, reminds me to that
I don't think there is a good mechanical analogue for this, although there were a lot of suggestions that came close. Phil's zoetrope seemed to be the best to me. Personally I think of it as a 16 position wafer switch with connections to 16 memory blocks from the 16 cogs. A round cross point switch.
Good ideas, Everyone. That thing Phil showed is an apt analogy, but almost nobody knows the name of that thing. I think the planetary gear is a good visual analogy, with maybe a central gear that is the hub ram, surrounded by 16 small gears.
I'll have to be the lone dissenter here I guess.
Trying to force some sort of car/plane/gear analogy as an explanation of what is happening seems to be the same old-same old as with the P1 before.
Not sure it helped pique any real substantial interest, regardless of how many of the converted swear its helped them explain the P1 to others.
Why not try something different this time, and use a more tech-familiar explanation based upon timing, ports, memory, cores, processes, threads, etc ?
No, I don't have one at the moment, however I'd still like to see Parallax focus explaining the new P2 in a way that does not remind anyone of propellers, goofy beanies, and cogs and gears.... :nerd:
Why not try something different this time, and use a more tech-familiar explanation based upon timing, ports, memory, cores, processes, threads, etc ?
If you really want to avoid anything circular as an illustrator, then that leaves a 16x16 crosspoint square, shown 2+ times to show the dynamic movement (rotate?) of all the crosspoints with each Sys Clock.
16 clocks later, it repeats. Always each HUB-memory-nibble connects to one COG.
or expressed as an eqn, something like
with 4 bit variabes, N, ClkCtr
FOREACH N, 1..16, COG[N] <= HUBsector[N+ClkCtr]
You are not alone. There is no really good analogy here, but my visualization is:
An outer circle or hexadecagon with 16 groups of 4 pins to represent the I/O bus and I/O pins.
Immediately inside this would be 16 wedges (or squares) to represent the cores.
Each core would have a tab facing in to represent the connection to the hub.
The hub would be 16 pie slices, each representing 32K of ram + whatever else is in the hub.
Each slice would have a tab facing out to the cores to represent the hub connection to the cores.
It would be the hub that rotates in the diagram.
The word "zoetrope" has a Greek etymology: zoe (life, animation) + trope (wheel). This could be used as the basis for something closer to home, e.g.
metatrope meta (data) + trope (wheel), or mnemotrope mnemo (memory) + trope (wheel).
We could also go the Latin route:
pocarota po (ponere to put) + ca (capere to get) + rota (wheel)
i.e. a wheel for puttting and getting data. "You do the pocarota, and you turn yourself around. That's what it's all about!"
-Phil
Addendum: Like the Chevy Nova ("no va" means "doesn't go"), pocarota doens't play well in Spanish: "poca rota" means "a little broken." Oh, well.
A descriptive name in Greek, Latin, or any other etymology would be fine, but most folks would probably have to google it for a definition. I certainly had to look up zoetrope. Being more of a visual rather than verbal thinker I am definitely in the "a picture is worth a thousand words" camp.
Note how the distinction between the old, like the Prop1, and new mechanics can be interpreted as a single vs multiple persons at the table. (Applies to all our examples given so far I think, ie: Substitute a person for a sprocket for example.)
Trying to force some sort of car/plane/gear analogy as an explanation of what is happening seems to be the same old-same old as with the P1 before.
Not sure it helped pique any real substantial interest, regardless of how many of the converted swear its helped them explain the P1 to others.
Why not try something different this time, and use a more tech-familiar explanation based upon timing, ports, memory, cores, processes, threads, etc ?
No, I don't have one at the moment, however I'd still like to see Parallax focus explaining the new P2 in a way that does not remind anyone of propellers, goofy beanies, and cogs and gears.... :nerd:
You are not the lone dissenter koehler, I agree with you.
How would Microchip, TI, Atmel and the others approach this?
Would they offer it up to a public forum to come up with a clever/cute descriptive name? I think not.
No, they'd give it to a marketing professional to come up with a technical description that would appeal to the professional engineers out there.
That is what Parallax will need to do if they are wanting to pitch this as a professional choice in MCU land.
It might be the boring option but anything else diminishes it's credibility in my opinion.
Comments
Marty
Don't know how long ---- But still there!
I'm glad you're still here, Sapieha!
Good ideas, Everyone. That thing Phil showed is an apt analogy, but almost nobody knows the name of that thing. I think the planetary gear is a good visual analogy, with maybe a central gear that is the hub ram, surrounded by 16 small gears.
We had a meeting with OnSemi today at Parallax to update where we're all at, since we've embarked on this redesign. The path to completion is simpler than ever, but there's still a good amount of Verilog work to do. Hopefully, I'll have an FPGA image soon.
Yeah, I know, I've got the P1 image contest to kick off, and I'll do it this week I promise, but this idea of the gears could use some love while we all get through waiting for the FPGA.
DE0 and DE2 ready for action....
Is the COG/HUB at a stage where OnSemi can run simulations on MHz and mA/MHz yet ?
But, there's no easy way to do that. 16 is a big number.
Now that I think about it some more, I think Phil's zoetrope is perhaps easier to visualize...
That means something like an animated crosspoint switch, or some approaching '70s "string art" might be the most applicable. We can put animations in pdfs now, right?
I did look for a 16 partner round dance but didn't succeed. The propeller do si do...
[ http://forums.parallax.com/attachment.php?attachmentid=108689&d=1400011603 ]
A Nibble sequential crosspoint is what the hardware does, which can be drawn as outers as COGs and inner as HUB memory, segmented into 16 Nibble fractions.
Minor tweaks to Chips drawing I'd do would be
* To merge/expand the yellow HUB memory (currently 16 isolated dots), to make it clearer it is not 16 smaller memories to the user.
* Present 2 (or more) copies to underscore the Rotating Physical model eg one @Cyc=0 and one @Cyc= +5
While looking at that drawing, I was reminded of a rotary-dial telephone. Do you think we could exploit that familiar concept? We'd just have to mention that each dial pulse has a 5ns cycle, instead of ~200ms. And, there are some other differences.
a) very few are going to know what a rotary-dial telephone looks like, or how it worked.
b) rotary-dial telephone have a single-pulse stream, the actual 16 way Mux connects all 16 COGS to one memory each, so it is important to preserve that in-parallel concept. The original image was very close.
Just show it in 2 positions to 'drive home' the (virtual) rotating cross point.
The other day, a friend of mine was visiting with his 14-year-old daughter, I have a rotary dial phone in my living room -- mostly as a conversation piece -- but it's still hooked up. My friend's daughter asked, "Papa, how do you work one of these?" He had to demonstrate it for her. I asked her if she'd ever played a phonograph record. "Nope." The times, they are a-changin'.
OTOH, a dial telephone is a "little" more up-to-date than a zoetrope!
-Phil
Good one!
I also like the rotary phone metaphor as well since I still have a rotary phone at home. Why use new when old tech still does the job?
Prior to our dial telephone we had a telephone that responded to voice commands. You picked up the receiver and waited to be acknowledged by the operator, Blanche. You could give Blanche the number you wanted or you could just tell her who you wanted to speak with. The phone numbers were only three digits but 'Nan Hapgood' or 'Aunt Stella' were easier for a four year old to remember. Blanche had excellent voice recognition skills so she knew who you were when you spoke and who 'Nan Hapgood' or 'Aunt Stella' were based on recognizing your voice. Kind of a Star Trek thing before its time.
In the evenings Blanche might not be home so sometimes you would have to wait until morning to make a call. No big deal.
Sandy
Zoetrochip?
Might be hard to spoof that the way Atmel did the Propeller: http://www.eetimes.com/document.asp?doc_id=1280547
I'll have to be the lone dissenter here I guess.
Trying to force some sort of car/plane/gear analogy as an explanation of what is happening seems to be the same old-same old as with the P1 before.
Not sure it helped pique any real substantial interest, regardless of how many of the converted swear its helped them explain the P1 to others.
Why not try something different this time, and use a more tech-familiar explanation based upon timing, ports, memory, cores, processes, threads, etc ?
No, I don't have one at the moment, however I'd still like to see Parallax focus explaining the new P2 in a way that does not remind anyone of propellers, goofy beanies, and cogs and gears.... :nerd:
If you really want to avoid anything circular as an illustrator, then that leaves a 16x16 crosspoint square, shown 2+ times to show the dynamic movement (rotate?) of all the crosspoints with each Sys Clock.
16 clocks later, it repeats. Always each HUB-memory-nibble connects to one COG.
or expressed as an eqn, something like
with 4 bit variabes, N, ClkCtr
FOREACH N, 1..16, COG[N] <= HUBsector[N+ClkCtr]
You are not alone. There is no really good analogy here, but my visualization is:
An outer circle or hexadecagon with 16 groups of 4 pins to represent the I/O bus and I/O pins.
Immediately inside this would be 16 wedges (or squares) to represent the cores.
Each core would have a tab facing in to represent the connection to the hub.
The hub would be 16 pie slices, each representing 32K of ram + whatever else is in the hub.
Each slice would have a tab facing out to the cores to represent the hub connection to the cores.
It would be the hub that rotates in the diagram.
mnemotrope mnemo (memory) + trope (wheel).
We could also go the Latin route:
i.e. a wheel for puttting and getting data. "You do the pocarota, and you turn yourself around. That's what it's all about!"
-Phil
Addendum: Like the Chevy Nova ("no va" means "doesn't go"), pocarota doens't play well in Spanish: "poca rota" means "a little broken." Oh, well.
A descriptive name in Greek, Latin, or any other etymology would be fine, but most folks would probably have to google it for a definition. I certainly had to look up zoetrope. Being more of a visual rather than verbal thinker I am definitely in the "a picture is worth a thousand words" camp.
Note how the distinction between the old, like the Prop1, and new mechanics can be interpreted as a single vs multiple persons at the table. (Applies to all our examples given so far I think, ie: Substitute a person for a sprocket for example.)
You are not the lone dissenter koehler, I agree with you.
How would Microchip, TI, Atmel and the others approach this?
Would they offer it up to a public forum to come up with a clever/cute descriptive name? I think not.
No, they'd give it to a marketing professional to come up with a technical description that would appeal to the professional engineers out there.
That is what Parallax will need to do if they are wanting to pitch this as a professional choice in MCU land.
It might be the boring option but anything else diminishes it's credibility in my opinion.