I just got the files done for the BeMicroCV board.
<snip>
It turns out that even though this Cyclone V -A2 part has 25k LE's, the P8X32A takes 85% of them! That's 33% more LE's than it takes on the Cyclone IV. Plus, it took 3x longer to compile. Oh, and Fmax is still over 80MHz, but slower than Cyclone IV. Only good thing is, these Cyclone V FPGA's cost half what the Cyclone IV devices cost in similar (higher) densities.
Thanks Chip, this is useful information.
The BeMicro board looks rather inexpensive and interesting, and its small SOIC i2c serial ROM looks more amenable to hand desoldering/replacement with a larger capacity part to emulate a regular prop setup with 32k8 bits of EEPROM when compared to the DE-0 Nano variant (whose device looks positively tiny in comparison). The SD card reader is a good addition too. But if it won't leave enough room for an SDRAM memory controller with the remaining logic after a prop is put in, it could be a bit of a show stopper (for me at least).
The DE-0 nano board seems a very nice fit except for the SRAM size, although perhaps that 16kB hole could be exploited to window into the SDRAM??? I keep looking at the second 40 pin header with interest for a 32 bit Port B implementation. Both boards have that ability which is really useful. And allocating a hardware 16x16 multiplier to each COG (actually 18x18 by the looks of it) and tying its use into the original instruction opcodes allocated for that purpose would be great for 16 bit audio processing too.
There is also a 0.1" row offset. If using the parallax de0 breakout the easiest thing might be to buy a 40 pin extended header like adafruit now sell (samtec also make these, they're common for PC104 boards), to separate the two boards and only connect the correct header through.
Just remember per Guy Lemieux's warning, only one of the 40 pin headers on the BeMicro is compatible with the DE0.
That's true, and it's only 28 GPIO pins on that second 40 pin header, not all 32. The remaining 4 port bits could potentially be mapped to the onboard SD card to get the full usage of all 32 bits available on a prop port, or alternatively included on the extra GPIO pins on the other expansion port for example. Lots of ways to make things work out nice.
It turns out that even though this Cyclone V -A2 part has 25k LE's, the P8X32A takes 85% of them! That's 33% more LE's than it takes on the Cyclone IV.
It is because these LEs are fake. It has new 4bit ALMs instead of 3bit LEs. They counts 1ALM=2.5 LE but it is simply not true. 1ALM=something slightly more than 2 LEs.Maybe 2.1. My C5 compilation took about 7900 ALMs...
It is because these LEs are fake. It has new 4bit ALMs instead of 3bit LEs. They counts 1ALM=2.5 LE but it is simply not true. 1ALM=something slightly more than 2 LEs.Maybe 2.1. My C5 compilation took about 7900 ALMs...
That makes sense.
We used to know a guy who had worked at Altera for a long time. He said that they develop new fabric topologies by running thousands of stable designs through them to weigh their relative merits. For some reason, when they went to 28nm they decided on an architecture that exploits the density, but not the speed potential. Cyclone V is slower than Cyclone IV and takes waaaaaaaay longer to compile, perhaps because there are more trade offs in mapping designs to the new ALMs than there were to the old LE's.
Will 64K of hub memory fit with a tiny loader ROM?
I cannot quite remember the precise hub rom layout. IIRC the booter and interpreter live right at the top of hub. So my guess is that you can get ~60KB of ram without doing any major changes.
This is Propeller land, there's always something fun to do!!! "The FUN never sets on the Propeller Empire!" (My appologies to our British friends)
Or you could install the Quartus tools which isn't really that much fun but it takes a while and better now when your new toy is in flight than after it lands and you want to play!
Start studying up on Verilog?
Start looking at the Verilog sources?
Knock items off the "Honey Do" list so you don't feel guilty when it's play time again?
My high score is only 31,488 so I feel I still must have a life outside of 2048!! I guess my next goal is actually 4096........it's good to have goals!!
No, my all time favorite is "Railroad Tycoon", I like to build empires. I cannot seem to find a version that is availble anymore. The last version that I did have was for a Linux, but that was many, many years ago.
My high score is only 31,488 so I feel I still must have a life outside of 2048!! I guess my next goal is actually 4096........it's good to have goals!!
4096 puts the score over 50,000 and the tile is white on black. I'll never make to 8192. It's a very interesting game for programmers passing the time waiting in line, etc....
I rebuilt the .jic file on my PC and thought I'd try blinking some LEDs instead of something needing serial I/O. I loaded my .jic, hooked up the LED and loaded a blinky program and it works!! I still show the COG LEDs (see picture below) but it does run Spin!
I reloaded Chip's .jic and it works with the blinky LEDs. OK, so I tried the PST Demo that comes with the Propeller Tool and it works.
Not sure what happened but it looks like the .jic in your .zip works!!
Now the remaining question, "Is there really a dead COG or is something else causing the LED not to be lit?"
Sorry for any confusion and the premature reporting.
====== back to the original message ======
Seeing that you already had provided a .jic file, I proceeded to load that rather than rebuilding on my system.
On recycle of power, it looked like this:
as you can see, I have 7 LEDs lit - the 3rd from the bottom isn't lit but at this point, I should just have one lit.
The Propeller tool was able to talk to it - F7 worked - I tried F10 and it said it worked but the program didn't run. It was the PST demo program and I didn't get any terminal output.
The PropPlug is actually on the correct pins - there are two empty pins below it, you just can't tell in the picture.
That's all I've done with it so far.
I'll try a local rebuild and some more experimenting later this afternoon.
Start off with the DE2-115 directory setup and get rid of all files that have DE2_115 in their names. Then, put these files into that folder and rename the folder P8X32A_BeMicroCV. After that you can follow directions in the readme.txt file.
It turns out that even though this Cyclone V -A2 part has 25k LE's, the P8X32A takes 85% of them! That's 33% more LE's than it takes on the Cyclone IV. Plus, it took 3x longer to compile. Oh, and Fmax is still over 80MHz, but slower than Cyclone IV. Only good thing is, these Cyclone V FPGA's cost half what the Cyclone IV devices cost in similar (higher) densities.
Maybe someone can try these files out and see if they work.
No worries. This British friend is English. Many of us English would like independence from the British Empire as well. At least give me a passport that does actually mention the name of the country I come from!
Hopefully the Scottish will get their independence soon and England can come back into existence as well.
Anyway, yes, Parallaxia is and endlessly fascinating land.
Got back from my road trip, all excited to order one of these and join the fun, only to find them sold out! I KNEW I should have ordered one before I left town.
Hopefully they had some in the pipe so it won't be too long before we can start ordering again.
Only problem is the board cost is $39.90 with a $60.00 shipping fee from Japan LOL. For that much I may just get a DE-Nano from Parallax... Wish their FPGA board was ready, or even a list of planned specs and target price, I would just wait for it.
Only problem is the board cost is $39.90 with a $60.00 shipping fee from Japan LOL. For that much I may just get a DE-Nano from Parallax... Wish their FPGA board was ready, or even a list of planned specs and target price, I would just wait for it.
Perhaps a group order would be worth it? Price is good at 10 off, wonder whether the $60 shipping goes up much for 10 units?
EDIT: Interestingly the price comes down to $33.40 if you order 25 (once in the basket the 25+ option appears). That's a pretty amazing price for what's packed onto that board. Since they have 19 they may give you all of those at the good price. Shipping to AU is about half the US fwiw, can be free (though I think the exchange rate makes this a worse option).
Its hard to know how long arrow north america will be out of stock for. In the past these have been out of stock for a while.
I just compiled the original files from Chip, the ones posted on Wed... it took ages to compile !... I am also on a Cyclone V, the BeMicro. I'll check these new files of him to see if a top entity is there... Note: Those "top" files have it all
I just skimped over his code... that is quite an art work !. My version is so primitive in comparison... , I have so a road ahead...
[about chip1stop] On their About page - but I can't verify that right now because their site is down for maintenance. I checked verical and arrow now and as it's not mentioned on their sites that only leaves chip1stop.. because I saw it while browsing for BEMicro CV.
I just compiled the original files from Chip, the ones posted on Wed... it took ages to compile !... I am also on a Cyclone V, the BeMicro. I'll check these new files of him to see if a top entity is there... Note: Those "top" files have it all
All things considered, I think the DE0-Nano is a better deal. It's a little more expensive and lacks the RAM needed for the character ROM, but it has a faster Cyclone IV chip with a lot more extra logic, and it compiles way faster. Plus, it's easier to get.
I got back from DEFCON last night to find a box with the BeMicroCV in it. I'll get the files tuned up for it tomorrow morning. I noticed that the LED's must be driven low to turn on, whereas the configuration I posted drives them high. That's why all but one are one after a reset. I don't know why they are out of order, though. I just double-checked the pin numbers and they were fine.
DEFCON was really fun. The spirit there was very much aligned with that of the forums here. It felt so much better than some trade show. Everybody there just enjoyed learning and sharing ideas. All the presenters were very down-to-earth. It's a big shift in paradigm that has really antiquated the corporate model.
All things considered, I think the DE0-Nano is a better deal. It's a little more expensive and lacks the RAM needed for the character ROM, but it has a faster Cyclone IV chip with a lot more extra logic, and it compiles way faster. Plus, it's easier to get.
The DE-0 nano looks like a really nice product. It appears to be almost a perfect fit for the Prop 1 design apart from the missing 16kB if you need the font memory, though that very hole could be useful for mapping in other FPGA peripherals. I've been wondering if there is a good way to go that could store the font data and be accessible within the 7 CPU clock cycle budget for hub reads. It may just able be possible to use the SDRAM for returning random access data (as you probably get 14 clocks if the SDRAM gets clocked at 2x the CPU), but if primarily designed for that purpose it could then get tricky to use it for other applications as well such as graphics buffers. There are 16 useful looking pins (3 read only) coming out on the bottom of the board. I wonder if there is a way to get a very fast RAM/ROM onto a daughterboard for this purpose using the pins/clocks available...hmm. Maybe 2 address latches + SRAM and a small FPGA state machine. 8 address/data pins, ALE1, ALE2, RD, WR and a CS might just about do it. That keeps the two 32 bit ports on top freed up for portA and portB. The bootup code would need a way to populate this memory though unless its a ROM (but they tend to be a bit slower).
[about chip1stop] On their About page - but I can't verify that right now because their site is down for maintenance. I checked verical and arrow now and as it's not mentioned on their sites that only leaves chip1stop.. because I saw it while browsing for BEMicro CV.
Comments
Thanks Chip, this is useful information.
The BeMicro board looks rather inexpensive and interesting, and its small SOIC i2c serial ROM looks more amenable to hand desoldering/replacement with a larger capacity part to emulate a regular prop setup with 32k8 bits of EEPROM when compared to the DE-0 Nano variant (whose device looks positively tiny in comparison). The SD card reader is a good addition too. But if it won't leave enough room for an SDRAM memory controller with the remaining logic after a prop is put in, it could be a bit of a show stopper (for me at least).
The DE-0 nano board seems a very nice fit except for the SRAM size, although perhaps that 16kB hole could be exploited to window into the SDRAM??? I keep looking at the second 40 pin header with interest for a 32 bit Port B implementation. Both boards have that ability which is really useful. And allocating a hardware 16x16 multiplier to each COG (actually 18x18 by the looks of it) and tying its use into the original instruction opcodes allocated for that purpose would be great for 16 bit audio processing too.
Happy days ahead.
Just remember per Guy Lemieux's warning, only one of the 40 pin headers on the BeMicro is compatible with the DE0.
http://forums.parallax.com/showthread.php/150849-BeMicro-CV-FPGA-Board-for-P2?p=1222356&viewfull=1#post1222356
There is also a 0.1" row offset. If using the parallax de0 breakout the easiest thing might be to buy a 40 pin extended header like adafruit now sell (samtec also make these, they're common for PC104 boards), to separate the two boards and only connect the correct header through.
http://forums.parallax.com/showthread.php/150849-BeMicro-CV-FPGA-Board-for-P2?p=1222323&viewfull=1#post1222323
That's true, and it's only 28 GPIO pins on that second 40 pin header, not all 32. The remaining 4 port bits could potentially be mapped to the onboard SD card to get the full usage of all 32 bits available on a prop port, or alternatively included on the extra GPIO pins on the other expansion port for example. Lots of ways to make things work out nice.
Rogloh if you're interested in getting the DDR3 up and running perhaps it would be possible to drop out a few cogs to fit a a DDR controller.
It is because these LEs are fake. It has new 4bit ALMs instead of 3bit LEs. They counts 1ALM=2.5 LE but it is simply not true. 1ALM=something slightly more than 2 LEs.Maybe 2.1. My C5 compilation took about 7900 ALMs...
That makes sense.
We used to know a guy who had worked at Altera for a long time. He said that they develop new fabric topologies by running thousands of stable designs through them to weigh their relative merits. For some reason, when they went to 28nm they decided on an architecture that exploits the density, but not the speed potential. Cyclone V is slower than Cyclone IV and takes waaaaaaaay longer to compile, perhaps because there are more trade offs in mapping designs to the new ALMs than there were to the old LE's.
Maybe but mechanics don't always have the best running cars LOL
And, the possibility for lots of extras... I'm suddenly more interested in this whole FPGA thing...
Ray
Or you could install the Quartus tools which isn't really that much fun but it takes a while and better now when your new toy is in flight than after it lands and you want to play!
Start studying up on Verilog?
Start looking at the Verilog sources?
Knock items off the "Honey Do" list so you don't feel guilty when it's play time again?
Play 2048 of course. http://gabrielecirulli.github.io/2048/
Ouch!! Just don't put it on your iPhone!
My high score is only 31,488 so I feel I still must have a life outside of 2048!! I guess my next goal is actually 4096........it's good to have goals!!
Ray
4096 puts the score over 50,000 and the tile is white on black. I'll never make to 8192. It's a very interesting game for programmers passing the time waiting in line, etc....
@Ray,
You might enjoy Lemonade Tycoon.
I just followed your instructions below.
===============
EDIT: NEWS FLASH
I rebuilt the .jic file on my PC and thought I'd try blinking some LEDs instead of something needing serial I/O. I loaded my .jic, hooked up the LED and loaded a blinky program and it works!! I still show the COG LEDs (see picture below) but it does run Spin!
I reloaded Chip's .jic and it works with the blinky LEDs. OK, so I tried the PST Demo that comes with the Propeller Tool and it works.
Not sure what happened but it looks like the .jic in your .zip works!!
Now the remaining question, "Is there really a dead COG or is something else causing the LED not to be lit?"
Sorry for any confusion and the premature reporting.
====== back to the original message ======
Seeing that you already had provided a .jic file, I proceeded to load that rather than rebuilding on my system.
On recycle of power, it looked like this:
as you can see, I have 7 LEDs lit - the 3rd from the bottom isn't lit but at this point, I should just have one lit.
The Propeller tool was able to talk to it - F7 worked - I tried F10 and it said it worked but the program didn't run. It was the PST demo program and I didn't get any terminal output.
The PropPlug is actually on the correct pins - there are two empty pins below it, you just can't tell in the picture.
That's all I've done with it so far.
I'll try a local rebuild and some more experimenting later this afternoon.
We're closer than we were this time yesterday!!
Hopefully the Scottish will get their independence soon and England can come back into existence as well.
Anyway, yes, Parallaxia is and endlessly fascinating land.
Ray
Hopefully they had some in the pipe so it won't be too long before we can start ordering again.
http://www.chip1stop.com/web/GBR/en/dispDetail.do?partId=ARRD-0000004&utm_medium=cpc&utm_campaign=aggregator&utm_source=fc&utm_term=ARRD-0000004&cid=fc_ARRD-0000004
They are an Arrow company, like Verical.
Thanks Leon!
Only problem is the board cost is $39.90 with a $60.00 shipping fee from Japan LOL. For that much I may just get a DE-Nano from Parallax... Wish their FPGA board was ready, or even a list of planned specs and target price, I would just wait for it.
Perhaps a group order would be worth it? Price is good at 10 off, wonder whether the $60 shipping goes up much for 10 units?
EDIT: Interestingly the price comes down to $33.40 if you order 25 (once in the basket the 25+ option appears). That's a pretty amazing price for what's packed onto that board. Since they have 19 they may give you all of those at the good price. Shipping to AU is about half the US fwiw, can be free (though I think the exchange rate makes this a worse option).
Its hard to know how long arrow north america will be out of stock for. In the past these have been out of stock for a while.
I tried Verical... can't find it there. Otherwise, the cupboard appears to be bare:)
Rich
I just skimped over his code... that is quite an art work !. My version is so primitive in comparison...
All things considered, I think the DE0-Nano is a better deal. It's a little more expensive and lacks the RAM needed for the character ROM, but it has a faster Cyclone IV chip with a lot more extra logic, and it compiles way faster. Plus, it's easier to get.
I got back from DEFCON last night to find a box with the BeMicroCV in it. I'll get the files tuned up for it tomorrow morning. I noticed that the LED's must be driven low to turn on, whereas the configuration I posted drives them high. That's why all but one are one after a reset. I don't know why they are out of order, though. I just double-checked the pin numbers and they were fine.
DEFCON was really fun. The spirit there was very much aligned with that of the forums here. It felt so much better than some trade show. Everybody there just enjoyed learning and sharing ideas. All the presenters were very down-to-earth. It's a big shift in paradigm that has really antiquated the corporate model.
The DE-0 nano looks like a really nice product. It appears to be almost a perfect fit for the Prop 1 design apart from the missing 16kB if you need the font memory, though that very hole could be useful for mapping in other FPGA peripherals. I've been wondering if there is a good way to go that could store the font data and be accessible within the 7 CPU clock cycle budget for hub reads. It may just able be possible to use the SDRAM for returning random access data (as you probably get 14 clocks if the SDRAM gets clocked at 2x the CPU), but if primarily designed for that purpose it could then get tricky to use it for other applications as well such as graphics buffers. There are 16 useful looking pins (3 read only) coming out on the bottom of the board. I wonder if there is a way to get a very fast RAM/ROM onto a daughterboard for this purpose using the pins/clocks available...hmm. Maybe 2 address latches + SRAM and a small FPGA state machine. 8 address/data pins, ALE1, ALE2, RD, WR and a CS might just about do it. That keeps the two 32 bit ports on top freed up for portA and portB. The bootup code would need a way to populate this memory though unless its a ROM (but they tend to be a bit slower).
I couldn't see it on their About page.