Who Will be Prop2's Biggest Competition?

135

Comments

  • Yes, there was a delay slot at jumps, and a jump took 4 cycles if I remember correct.

    You can't read the instruction and the register data in the same cycle, when you read an instruction the S and D values are from the previus instruction and the PC is already on the next instruction.
  • thej wrote: »
    cgracey wrote: »
    To get those three read ports and one write port, P2-Hot used THREE instances of 512x32 dual-port RAM per cog.

    The current P2 uses just one instance of 512x32 dual-port RAM per cog.

    Did the P2-Hot use any pipelining?
    J

    Yes. Three stages.
  • It seems the conversation's gone off topic...
  • Actually the conversation is on topic when you realize that the biggest competitors to the P2 are the various versions of the P2 itself. All of these variations have prevented the P2 from coming out in a timely manner. This has caused the P2 to fall behind other chips that continue to evolve in the market.

    I admit that many of the changes have been necessary, but it also must be admitted that many of the changes were arbitrary and not necessary. Even within the past few days people are proposing changes to bit fields and instruction encodings that aren't necessary.

    The P2 will serve well as a replacement for the P1 where higher speed, more memory and more I/O pins are needed. People who already use the P1 will probably use the P2 as well. People who have used the P1 in the past, but have moved on to other processors may return to use the P2. Parallax will get new customers for the P2, but the customer base will be limited because of the lack of tools.

    If Parallax is really serious about selling the P2 to new customers they should have been working on the development tools that will be needed. I understand that Parallax has limited resources. However, it seems like they could have invested $100K/year to pay someone to lead a group of volunteers to develop the GCC tools for the P2. I guess the strategy is to wait until the P2 chip is a reality before starting this effort. However, that delays the point where a fully supported P2 is available by another year or two. So in some sense, this strategy is causing the P2 to be less competitive.
  • David BetzDavid Betz Posts: 11,683
    edited January 4 Vote Up0Vote Down
    Dave Hein wrote: »
    Actually the conversation is on topic when you realize that the biggest competitors to the P2 are the various versions of the P2 itself. All of these variations have prevented the P2 from coming out in a timely manner. This has caused the P2 to fall behind other chips that continue to evolve in the market.

    I admit that many of the changes have been necessary, but it also must be admitted that many of the changes were arbitrary and not necessary. Even within the past few days people are proposing changes to bit fields and instruction encodings that aren't necessary.

    The P2 will serve well as a replacement for the P1 where higher speed, more memory and more I/O pins are needed. People who already use the P1 will probably use the P2 as well. People who have used the P1 in the past, but have moved on to other processors may return to use the P2. Parallax will get new customers for the P2, but the customer base will be limited because of the lack of tools.

    If Parallax is really serious about selling the P2 to new customers they should have been working on the development tools that will be needed. I understand that Parallax has limited resources. However, it seems like they could have invested $100K/year to pay someone to lead a group of volunteers to develop the GCC tools for the P2. I guess the strategy is to wait until the P2 chip is a reality before starting this effort. However, that delays the point where a fully supported P2 is available by another year or two. So in some sense, this strategy is causing the P2 to be less competitive.
    I think some volunteer effort might surface for working on PropGCC for the P2 when it becomes clear that the design is really frozen. I know that statement has been made recently and I suspect it's probably true. However, P2-hot was also frozen but when they tried to synthesize it it failed. Maybe once OnSemi gets the current P2 through synthesis and has a reasonable hope that the design is feasible then efforts may pick up.

  • jmgjmg Posts: 10,760
    Dave Hein wrote: »
    Actually the conversation is on topic when you realize that the biggest competitors to the P2 are the various versions of the P2 itself. All of these variations have prevented the P2 from coming out in a timely manner. This has caused the P2 to fall behind other chips that continue to evolve in the market.
    Hehe, that is one way of looking at it...


    David Betz wrote: »
    I think some volunteer effort might surface for working on PropGCC for the P2 when it becomes clear that the design is really frozen. I know that statement has been made recently and I suspect it's probably true. However, P2-hot was also frozen but when they tried to synthesize it it failed. Maybe once OnSemi gets the current P2 through synthesis and has a reasonable hope that the design is feasible then efforts may pick up.

    Chip has passed the Verilog to OnSemi, so that is a milestone, as they start the scripts to steer all of this to a usable result. Chip has said they can accept late changes/fixes to the verilog, but nothing majour.

    Seems this flow is rather similar to PCB design, where you can chuck everything at an autorouter, but get a MHz and kB-RAM result that is below par. Human intervention in the form of scripts, can improve the results every time, until some 'good enough' point is reached for the MHz and kB-RAM numbers.
    We should see some of those 'earliest pass' MHz and kB-RAM numbers quite soon, and OnSemi must have a feel as to what the improvement curve looks like, and thus where they might hope to end up.

    I'm still wary as to what the RAM numbers will come out at, as the logic has been quietly creeping, whilst the die size is fixed... concrete numbers will come soon.
    Fingers crosses the result is actually a viable device.
  • cgraceycgracey Posts: 8,538
    edited January 4 Vote Up0Vote Down
    jmg wrote: »
    ...I'm still wary as to what the RAM numbers will come out at, as the logic has been quietly creeping, whilst the die size is fixed... concrete numbers will come soon.
    Fingers crosses the result is actually a viable device.

    Good news yesterday from OnSemi:
    Chip,

    The # of instances for the core is at 723,625 vs when we last compiled at 722,480. So it didn't grow too much. I am still working through the details of the top level and how to best handle the I/O ring that you will be handing off as a GDS. When might we see a version of the I/O ring GDS?

    So, since they did their last full check, in order to quote us a complete price on their work, the logic has only grown by 0.16%. That's with all the changes from the past two months. I don't think there's much likelihood of any big surprise that would force us to reduce anything.
  • Is an 'instance' like a logic element in the FPGAs?
  • Tubular wrote: »
    Is an 'instance' like a logic element in the FPGAs?

    It could be a gate, a flipflop, or even a memory. These new instances are all gates and flipflops, not memories (which are big).
  • TonyB_TonyB_ Posts: 293
    edited January 5 Vote Up0Vote Down
    cgracey wrote: »
    jmg wrote: »
    ...I'm still wary as to what the RAM numbers will come out at, as the logic has been quietly creeping, whilst the die size is fixed... concrete numbers will come soon.
    Fingers crosses the result is actually a viable device.

    Good news yesterday from OnSemi:
    Chip,

    The # of instances for the core is at 723,625 vs when we last compiled at 722,480. So it didn't grow too much. I am still working through the details of the top level and how to best handle the I/O ring that you will be handing off as a GDS. When might we see a version of the I/O ring GDS?

    So, since they did their last full check, in order to quote us a complete price on their work, the logic has only grown by 0.16%. That's with all the changes from the past two months. I don't think there's much likelihood of any big surprise that would force us to reduce anything.

    Great news that the top-secret changes to XORO32 added so much for so little extra logic! (Only a small proportion of the +0.16% in reality, I guess.)

    Presumably OnSemi will implement adders as some variety of parallel prefix, which are much faster than carry lookahead and smaller according to this interesting master's degree report:

    https://repositories.lib.utexas.edu/bitstream/handle/2152/35326/DALMIYA-MASTERSREPORT-2015.pdf

    Formerly known as TonyB
  • TonyB_ wrote: »
    cgracey wrote: »
    jmg wrote: »
    ...I'm still wary as to what the RAM numbers will come out at, as the logic has been quietly creeping, whilst the die size is fixed... concrete numbers will come soon.
    Fingers crosses the result is actually a viable device.

    Good news yesterday from OnSemi:
    Chip,

    The # of instances for the core is at 723,625 vs when we last compiled at 722,480. So it didn't grow too much. I am still working through the details of the top level and how to best handle the I/O ring that you will be handing off as a GDS. When might we see a version of the I/O ring GDS?

    So, since they did their last full check, in order to quote us a complete price on their work, the logic has only grown by 0.16%. That's with all the changes from the past two months. I don't think there's much likelihood of any big surprise that would force us to reduce anything.

    Great news that the top-secret changes to XORO32 added so much for so little extra logic! (Only a small proportion of the +0.16% in reality, I guess.)

    Presumably OnSemi will implement adders as some variety of parallel prefix, which are much faster than carry lookahead and smaller according to this interesting master's degree report:

    https://repositories.lib.utexas.edu/bitstream/handle/2152/35326/DALMIYA-MASTERSREPORT-2015.pdf

    That was an interesting paper. The meat is all at the end.

    For synthesis, I'm pretty sure that OnSemi uses Synopsis Designware IP for all the logic, including adders and multipliers. I don't know what adder topologies are supported at 180nm, but I'd bet that at the smaller nodes, they support those fast adders mentioned in that paper.
  • Anyway, I'm still not aware of any other microcontroller that can handle more than two quadrature encoders and robotics is supposed to be the future, not to mention more and more CNC.
    PropBASIC ROCKS!
  • KeithE wrote: »
    There’s a guy “Gnarly Grey” selling UPDuino iCE40UP5K boards for $8. Are there many Boards that pair a microcontroller with a low cost FPGA? Parallax could probably make a nice educational product in this space if they thought it was worth it.
    This board sounded cool and also cheap so I ordered one. It came the other day and I'm disappointed about one thing. The RGB LED was not soldered to the board. You have to solder it yourself. Unfortunately, I'm not practiced at SMT soldering and I think the plastic on the RGB LED is very easy to melt. I don't understand why they didn't solder on the LED before shipping. Ugh.

  • jmgjmg Posts: 10,760
    David Betz wrote: »
    KeithE wrote: »
    There’s a guy “Gnarly Grey” selling UPDuino iCE40UP5K boards for $8. Are there many Boards that pair a microcontroller with a low cost FPGA? Parallax could probably make a nice educational product in this space if they thought it was worth it.
    This board sounded cool and also cheap so I ordered one. It came the other day and I'm disappointed about one thing. The RGB LED was not soldered to the board. You have to solder it yourself. Unfortunately, I'm not practiced at SMT soldering and I think the plastic on the RGB LED is very easy to melt. I don't understand why they didn't solder on the LED before shipping. Ugh.
    Perhaps they tried that, and had poor yields, so had to do the next best thing.... ? IIRC parallax had issues too with reflow profiles and LEDs...

    Looks like a gull wing package - not too hard, and you can pre-tin the LED leads to make things faster and thus cooler.

  • jmg wrote: »
    David Betz wrote: »
    KeithE wrote: »
    There’s a guy “Gnarly Grey” selling UPDuino iCE40UP5K boards for $8. Are there many Boards that pair a microcontroller with a low cost FPGA? Parallax could probably make a nice educational product in this space if they thought it was worth it.
    This board sounded cool and also cheap so I ordered one. It came the other day and I'm disappointed about one thing. The RGB LED was not soldered to the board. You have to solder it yourself. Unfortunately, I'm not practiced at SMT soldering and I think the plastic on the RGB LED is very easy to melt. I don't understand why they didn't solder on the LED before shipping. Ugh.
    Perhaps they tried that, and had poor yields, so had to do the next best thing.... ? IIRC parallax had issues too with reflow profiles and LEDs...

    Looks like a gull wing package - not too hard, and you can pre-tin the LED leads to make things faster and thus cooler.
    Well, if they had trouble soldering the LED, I'm sure I will have even worse luck. I'll try though.
  • David Betz wrote: »
    jmg wrote: »
    David Betz wrote: »
    KeithE wrote: »
    There’s a guy “Gnarly Grey” selling UPDuino iCE40UP5K boards for $8. Are there many Boards that pair a microcontroller with a low cost FPGA? Parallax could probably make a nice educational product in this space if they thought it was worth it.
    This board sounded cool and also cheap so I ordered one. It came the other day and I'm disappointed about one thing. The RGB LED was not soldered to the board. You have to solder it yourself. Unfortunately, I'm not practiced at SMT soldering and I think the plastic on the RGB LED is very easy to melt. I don't understand why they didn't solder on the LED before shipping. Ugh.
    Perhaps they tried that, and had poor yields, so had to do the next best thing.... ? IIRC parallax had issues too with reflow profiles and LEDs...

    Looks like a gull wing package - not too hard, and you can pre-tin the LED leads to make things faster and thus cooler.
    Well, if they had trouble soldering the LED, I'm sure I will have even worse luck. I'll try though.
    They aren't gull wing packages. The contacts don't extend beyond the plastic body of the LED. I have no idea how to solder it.
  • jmgjmg Posts: 10,760
    David Betz wrote: »
    Well, if they had trouble soldering the LED, I'm sure I will have even worse luck. I'll try though.
    The reflow issues are more temperature duration, and peak, and manual soldering usually is less on both.
    In reflow, everything has time to elevate to the peak temperature, and with IR lamp reflow, the parts can get even hotter....
    With manual soldering, only one lead-end is at solder temperature at a time. User lower temperature (lead) solder if you have any ?
  • jmg wrote: »
    David Betz wrote: »
    Well, if they had trouble soldering the LED, I'm sure I will have even worse luck. I'll try though.
    The reflow issues are more temperature duration, and peak, and manual soldering usually is less on both.
    In reflow, everything has time to elevate to the peak temperature, and with IR lamp reflow, the parts can get even hotter....
    With manual soldering, only one lead-end is at solder temperature at a time. User lower temperature (lead) solder if you have any ?
    Okay but how do you heat up a lead that is under the plastic housing of the LED without melting the plastic?

  • jmgjmg Posts: 10,760
    David Betz wrote: »
    Okay but how do you heat up a lead that is under the plastic housing of the LED without melting the plastic?

    Depends on the PCB pads and package. If you cannot access the pads, maybe change the LED ?

    Do you know the part code / package of the one they included ?

    The image here
    http://gnarlygrey.atspace.cc/development-platform.html#upduino
    Looks like one of these - SOJ PLCC4
    https://www.digikey.com/product-detail/en/broadcom-limited/ASMB-MTB0-0A3A2/516-3279-2-ND
    However, I see those are not cheapest RGB, - that one is ~17c.

    This one is new, lists at 9.3c/8k, also in gull wing PLCC4, but smaller at PLCC4 2220
    https://media.digikey.com/pdf/Data Sheets/Avago PDFs/ASMB-KTF0-0A306.pdf
    Shows stock due 14 Feb

    This one is larger PLCC4 3528
    https://www.digikey.com/product-detail/en/inolux/IN-P32TATRGB/1830-1116-1-ND
  • Was that led simply floating around in the package? If so then it had been soldered which seems logical but improperly of course and had come adrift during shipping. Those devices are not easy to solder by hand or even remove easily but you would certainly need solder paste and either reflow it or use a very good tip with sufficient heat or just forget it.
    Tachyon Forth - compact, fast, forthwright and interactive
    useforthlogo-s.png
    Tachyon Forth News Blog
    TACHYON DEMONSTRATOR
    Brisbane, Australia
  • Was that led simply floating around in the package? If so then it had been soldered which seems logical but improperly of course and had come adrift during shipping. Those devices are not easy to solder by hand or even remove easily but you would certainly need solder paste and either reflow it or use a very good tip with sufficient heat or just forget it.
    It wasn't just floating in the package. It was in a plastic carrier. It had never been installed.

  • David, I'd be happy to send you a solder challenge kit to sharpen your skills if that'd help
    https://hackaday.com/2016/06/19/soldering-challenge-to-challenge-you/

    Are you sure the metal doesn't go up the side of the led, even a little way? That can make a big difference. Even if you have just a few thou to latch onto. Its rare to not have something on the side, but I have seen a few

  • David,
    Just give it a go!

    Best if you have a flux pen and swipe a tiny amount on both the pads and underside of the LED.

    Melt a tiny blob of solder (use fine solder) onto your small soldering iron tip.
    Place the LED onto the board and while holding it in place, touch the corner of one exposed pcb pad. This should tack the LED to the board. Now you can solder the other 3 pads using fine solder wire and soldering iron. Just leave the soldering iron tip on the pcb long enough for the solder to melt on the pad. Careful - you do not want to heat the LED too much. Finally resolder the first pad.

    Small soldering iron with ~1.5mm fine tip. Temperature control is nice to have but not essential.
    Fine solder wire (sorry not home so cannot tell you the best size) with flux. Leaded is best because it melts as a lower temperature. Unleaded is better for the environment and you, but I am still here and I used leaded solder for decades.
    A flux pen helps heaps, but you can do it without this.

    The quality of the above depends on whether this will be a once off, or many times.
    My Prop boards: P8XBlade2, RamBlade, CpuBlade, TriBlade
    Prop OS (also see Sphinx, PropDos, PropCmd, Spinix)
    Website: www.clusos.com
    Prop Tools (Index) , Emulators (Index) , ZiCog (Z80)
  • Hey, solder leads to it, and then to the board. Stick it somewhere.
    Do not taunt Happy Fun Ball! @opengeekorg ---> Be Excellent To One Another SKYPE = acuity_doug
    Parallax colors simplified: http://forums.parallax.com/showthread.php?123709-Commented-Graphics_Demo.spin<br>
  • potatohead wrote: »
    Hey, solder leads to it, and then to the board. Stick it somewhere.
    I thought of that but I don't think I should have to do that. The product is misrepresented. It doesn't say anywhere that I could see that the LED had to be soldered on. If it had, I wouldn't have bought the board. Of course, it was only $8 so it isn't a big deal but it was disappointing. I sent email to the maker mentioning this and asking if there was a way to buy a board with the LED already soldered on.

  • Yeah, you shouldn't. It arrived broken, from your point of view.

    Do not taunt Happy Fun Ball! @opengeekorg ---> Be Excellent To One Another SKYPE = acuity_doug
    Parallax colors simplified: http://forums.parallax.com/showthread.php?123709-Commented-Graphics_Demo.spin<br>
  • David Betz wrote: »
    I thought of that but I don't think I should have to do that. The product is misrepresented. It doesn't say anywhere that I could see that the LED had to be soldered on. If it had, I wouldn't have bought the board. Of course, it was only $8 so it isn't a big deal but it was disappointing. I sent email to the maker mentioning this and asking if there was a way to buy a board with the LED already soldered on.

    Certainly that should not be their long term plan.

    At the very least, it has to be more costly to cut and insert single LEDS !!

    Sounds more like they had to do a quick fix, faced with something like yield issues.

    Does it look like yours had a LED removed ?
    Any part numbers or photos of the LED they did ship, as the photo I can see, sure looks like a gull wing PLCC4 package/
  • jmg wrote: »
    David Betz wrote: »
    I thought of that but I don't think I should have to do that. The product is misrepresented. It doesn't say anywhere that I could see that the LED had to be soldered on. If it had, I wouldn't have bought the board. Of course, it was only $8 so it isn't a big deal but it was disappointing. I sent email to the maker mentioning this and asking if there was a way to buy a board with the LED already soldered on.

    Certainly that should not be their long term plan.

    At the very least, it has to be more costly to cut and insert single LEDS !!

    Sounds more like they had to do a quick fix, faced with something like yield issues.

    Does it look like yours had a LED removed ?
    Any part numbers or photos of the LED they did ship, as the photo I can see, sure looks like a gull wing PLCC4 package/
    Maybe it is a gull wing package but I thought that would indicate leads that stick out from the side of the package. These leads look more like a Propeller QFN. I'll take a picture and post it later.

  • David Betz wrote: »
    Maybe it is a gull wing package but I thought that would indicate leads that stick out from the side of the package. These leads look more like a Propeller QFN. I'll take a picture and post it later.

    Sorry, I meant J lead, where the lead comes out the sides and bends under, like the PLCC44 packages. - hence the PLCC-4 package name.

    Most LEDs have side accessible leads, some are gold plated concave PCB edges, but there is a push to allow high density PCB packing, and leads out the sides are a pain then.

    After some searching, I did find a ROHM LED that has 6 terminals, completely under the package.
    http://www.rohm.com/web/global/datasheet/MSL0402RGBU

    To manually solder that, you would hot air gun, and need paste, or pre-tin and manage it like BGA where you do solder-solder reflow.

  • Fairly sure the pcb is the same as I bought last year. If so, I dont recall the RGB LED at all. If it was there it was fitted, else it wasnt present. Definately didnt solder anything. I havent used it though.
    My Prop boards: P8XBlade2, RamBlade, CpuBlade, TriBlade
    Prop OS (also see Sphinx, PropDos, PropCmd, Spinix)
    Website: www.clusos.com
    Prop Tools (Index) , Emulators (Index) , ZiCog (Z80)
Sign In or Register to comment.