Shop OBEX P1 Docs P2 Docs Learn Events
Prop2 FPGA files!!! - Updated 2 June 2018 - Final Version 32i - Page 64 — Parallax Forums

Prop2 FPGA files!!! - Updated 2 June 2018 - Final Version 32i

16162646667160

Comments

  • RaymanRayman Posts: 14,652
    Are the new "Bitmask/trim instructions" described somewhere?
  • cgraceycgracey Posts: 14,155
    Rayman wrote: »
    My P123 board seems to work like before after the update.

    Took me a moment to figure out Prop_Chk 0 0 0 0...
    Had to read instructions to remember that you have to put in a leading space.
    Had to set baud to 115200

    I'm also seeing that paste in PST doesn't work right.

    Does PNut now use this text serial mode to load code?
    Seems to load just as fast as before...

    Yes, PNut.exe uses this. It's all there is now. It's much easier to code for than what we had before, especially from high-level languages.

    Paste is broken in PST. Jeff is going to fix it.
  • So with the new FPGA image we should be able to adapt one of the P1 loaders to load a binary, correct? Is there a spec on the serial loader protocol? Is it similar to the P1 serial protocol? This will allow us to assemble code and load it independent from PNut, and will allow development on non-Windows platforms.
  • cgraceycgracey Posts: 14,155
    Dave Hein wrote: »
    So with the new FPGA image we should be able to adapt one of the P1 loaders to load a binary, correct? Is there a spec on the serial loader protocol? Is it similar to the P1 serial protocol? This will allow us to assemble code and load it independent from PNut, and will allow development on non-Windows platforms.

    At the end of the Google doc file, I detailed the whole protocol. It's very simple.
  • Dave HeinDave Hein Posts: 6,347
    edited 2016-09-11 20:48
    cgracey wrote: »
    Dave Hein wrote: »
    So with the new FPGA image we should be able to adapt one of the P1 loaders to load a binary, correct? Is there a spec on the serial loader protocol? Is it similar to the P1 serial protocol? This will allow us to assemble code and load it independent from PNut, and will allow development on non-Windows platforms.

    At the end of the Google doc file, I detailed the whole protocol. It's very simple.
    Great! Thanks. It looks like it's time for me to dust off the DE2-115 and try out the latest FPGA image.

  • I'm a bit confused on the change of the LUT sharing events from SETRDL/SETWRL to the SETSE* selectable-event paradigm. It must be a senior moment, but I'm not getting how a 9-bit LUT location plugs into the new syntax:

    %000_00_11AA = odd/even companion cog writes LUT address %1111111AA

    The other selectable-event setups make sense to me, and I can see that bits 3..2 of the above set the LUT event mode, but for some reason I'm hung up interpreting the "AA" part :confused:

    In p2v10a, to set up a write event from the other cog to this cogs' LUT was:
    SETLUT #1 ' Allow other cog to write this LUT
    SETWRL ##%1_000000000 | LUT_ADDR

    What would the equivalent syntax be using SETSE1?

    Thanks!
  • Rayman wrote: »
    Are the new "Bitmask/trim instructions" described somewhere?

    Here's what I found with the bit mask instructions
    BMASK D,S/# returns a bit mask of length S/#S + 1
    BMASK myreg,#3
     myreg = $f
    BMASK myreg,#5
     myreg = $3f
    
    
    BMASKN D,S/# returns a complementary bit mask of length S/#S + 1
    BMASKN myreg,#3
     myreg = $fffffff0
    BMASKN myreg,#5
     myreg = $ffffffc0
    
    
    BTRIM D,S/# trims a register to S + 1 bits
    mov myreg,##$ffffffff
    BTRIM myreg,#8
     myreg = $1ff
    
    
    
    
  • Chip
    Version 11b on Nano-bare working Ok now. :)
    Thanks
  • cgraceycgracey Posts: 14,155
    edited 2016-09-12 00:04
    garryj wrote: »
    I'm a bit confused on the change of the LUT sharing events from SETRDL/SETWRL to the SETSE* selectable-event paradigm. It must be a senior moment, but I'm not getting how a 9-bit LUT location plugs into the new syntax:

    %000_00_11AA = odd/even companion cog writes LUT address %1111111AA

    The other selectable-event setups make sense to me, and I can see that bits 3..2 of the above set the LUT event mode, but for some reason I'm hung up interpreting the "AA" part :confused:

    In p2v10a, to set up a write event from the other cog to this cogs' LUT was:
    SETLUT #1 ' Allow other cog to write this LUT
    SETWRL ##%1_000000000 | LUT_ADDR

    What would the equivalent syntax be using SETSE1?

    Thanks!

    You can select $1FC..$1FF for the LUT read or write address event sensitivity. That's what the AA is about.
  • I'm testing some of the new instructions too in Tachyon where I create an interactive version of them.
    TF2# .S
     DATA   x1    DEAD.BEEF
     BRANCH x0
     LOOP   x0
     RETURN x3    0000.26D8 0000.4FDE 0000.37EB ok
    TF2# 8 BMASK .LONG 0000.01FF ok
    TF2# 8 BMASKN .LONG FFFF.FE00 ok
    TF2# $12345678 8 BTRIM .LONG 0000.0078 ok
    TF2# $12345678 15 BTRIM .LONG 0000.5678 ok
    TF2#
    
  • cgracey wrote: »
    You can select $1FC..$1FF for the LUT read or write address event sensitivity. That's what the AA is about.
    Ah. Like I said -- a senior moment :blush:
    Thanks, Chip!
  • rjo__rjo__ Posts: 2,114
    Win10 P123-A9
    In the terminal:

    C:\Users\Public\p2\P2v11>C:\Users\Public\p2\P2v11\PX.exe C:\Users\Public\p2\P2v11\Prop123_A9_Prop2_v11.rbf /P/4

    returns "file /P/4 Not Found"

    I assume that this is related to the change in the boot loader?

    In the terminal:
    C:\Users\Public\p2\P2v11>C:\Users\Public\p2\P2v11\PX.exe C:\Users\Public\p2\P2v11\Prop123_A9_Prop2_v11.rbf /4

    Works fine.

    But I then flip the PGM/Run switch to RUN and cycle power, wait for the red 0 LED to light, and then PNUT doesn't find the P2.

    What am I doing wrong?



  • Try a space between "/P /4"
  • rjo__rjo__ Posts: 2,114
    oops. I don't need to cycle power... then we are good to go.

    sorry.
  • rjo__rjo__ Posts: 2,114
    edited 2016-09-14 16:53
    Chip,

    P2v11!!! love it. Updating my code was really a pleasure.


    setedg/waitedg have changed to setse1(2 or 3)/waitse1


    the compiler throws the expected error for the old instruction, setedg, but misses waitedg
  • rjo__rjo__ Posts: 2,114
    it looks for this bmp
  • cgraceycgracey Posts: 14,155
    rjo__ wrote: »
    Chip,

    P2v11!!! love it. Updating my code was really a pleasure.


    setedg/waitedg have changed to setse1(2 or 3)/waitse1


    the compiler throws the expected error for the old instruction, setedg, but misses waitedg

    It thinks WAITEDG is a label now, as it has no operands.
  • rjo__rjo__ Posts: 2,114
    K
  • cgraceycgracey Posts: 14,155
    I've always wanted to be able to call a subroutine and pass a parameter in the same instruction.

    We now have two instructions that do this:

    CALLP0 D/#,S/#
    CALLP1 D/#,S/#

    These instructions write D/# to either register $000 (CALLP0) or $001 (CALLP1) and call to S/#.

    D/# is a register or a constant.

    S/# is a register or a constant 9-bit relative address. So, if you are using a constant address, it must be within -$100 to +$FF instructions.

    These new instructions change this...

    MOV 0,#parameter
    CALL #subroutine

    ...to this...

    CALLP0 #parameter,#subroutine

    I made room for these instructions by replacing SETPAE/SETPAN/SETPBE/SETPBN with SETPEQ/SETPNE, where the C flag now selects the INA/INB port, instead of the instruction.

    These instructions are going to get the whole boot ROM back under 2KB. I made the last location of the ROM read first, so that the value can be DECOD'd to a 2^n byte count. This way, the whole 16KB ROM doesn't need to be loaded on reset. Only what is required gets loaded from ROM into hub, saving some milliseconds. In the future, the idea is that we can make the boot ROM up to 16KB, but never load more than is necessary after a reset.
  • jmgjmg Posts: 15,173
    cgracey wrote: »
    These instructions are going to get the whole boot ROM back under 2KB. I made the last location of the ROM read first, so that the value can be DECOD'd to a 2^n byte count. This way, the whole 16KB ROM doesn't need to be loaded on reset. Only what is required gets loaded from ROM into hub, saving some milliseconds. In the future, the idea is that we can make the boot ROM up to 16KB, but never load more than is necessary after a reset.

    Sounds great.
    The ROM is 16k Bytes or Bits ?
    Does the Boot code load into a COG, or simply stream into HUB and exec from there, meaning there is less of a hard boundary at 2KB.
    I'm guessing ROM is Reset-read only, and you cannot read-later ?

  • cgraceycgracey Posts: 14,155
    jmg wrote: »
    cgracey wrote: »
    These instructions are going to get the whole boot ROM back under 2KB. I made the last location of the ROM read first, so that the value can be DECOD'd to a 2^n byte count. This way, the whole 16KB ROM doesn't need to be loaded on reset. Only what is required gets loaded from ROM into hub, saving some milliseconds. In the future, the idea is that we can make the boot ROM up to 16KB, but never load more than is necessary after a reset.

    Sounds great.
    The ROM is 16k Bytes or Bits ?
    Does the Boot code load into a COG, or simply stream into HUB and exec from there, meaning there is less of a hard boundary at 2KB.
    I'm guessing ROM is Reset-read only, and you cannot read-later ?

    The ROM is 16K bytes. It gets streamed right into the hub, so that it can be executed. How much gets streamed into the hub is ROM-data-dependent. We will synthesize the whole 16KB ROM, but only load 2KB or 4KB, for now. Maybe someday P63 and P62 could be a USB pair with complete interface in ROM boot code, along with a fancy boot tree.

    The ROM can be read later, but it's prone to byte-skips if other cogs are doing the requisite 'CLKSET #0 WC' which reads the next byte of ROM into cog register $000.
  • cgracey wrote: »
    The ROM can be read later, but it's prone to byte-skips if other cogs are doing the requisite 'CLKSET #0 WC' which reads the next byte of ROM into cog register $000.

    I've been wondering about this. If you have a multi-stage boot, how would you be able to make use of the ROM authentication/encryption code from your first stage code for loading the next stage? I'm assuming the second-stage key would have to be embedded in the first-stage loader?

  • I like the Propeller and I am willing to donate one or two Configum modules to enable its miniaturization.
    https://www.tindie.com/products/KnivD/configum-fpga-module/

  • jmgjmg Posts: 15,173
    edited 2016-09-26 20:21
    knivd wrote: »
    I like the Propeller and I am willing to donate one or two Configum modules to enable its miniaturization.
    https://www.tindie.com/products/KnivD/configum-fpga-module/

    Nice board, & impressive tooling to match PLCC68.
    However, I'd worry a little about the long term supply of PLCC68 - we just dropped PLCC44 sockets, because their price climbed, and availability got more difficult on the brand we had approved.



    P2 builds do not yet target Xilinx flows, but the P1V has been rebuilt on Xilinx - you could take those builds, and check/tweak
    so they run on configum, & report back in a P1V thread ?

    Some P1V users, may like the PLCC68 form factor.

    Another recent Xilinx-PCB, with a similar price to your link is this one (DIP48 form factor, FT2232H included)
    http://store.digilentinc.com/cmod-a7-breadboardable-artix-7-fpga-module/

    The spec's are very thin on configum, can you do a simple resource table, similar to Digilent's ?
  • cgraceycgracey Posts: 14,155
    Seairth wrote: »
    cgracey wrote: »
    The ROM can be read later, but it's prone to byte-skips if other cogs are doing the requisite 'CLKSET #0 WC' which reads the next byte of ROM into cog register $000.

    I've been wondering about this. If you have a multi-stage boot, how would you be able to make use of the ROM authentication/encryption code from your first stage code for loading the next stage? I'm assuming the second-stage key would have to be embedded in the first-stage loader?

    The 2nd-stage booter, loaded from flash or serial, is $F8 instructions long, and loads into LUT $100.. $1F7. It is JMP'd to, with the 1st-stage loader still in other parts of the cog, along with the fuses.
  • Cluso99Cluso99 Posts: 18,069
    knivd wrote: »
    I like the Propeller and I am willing to donate one or two Configum modules to enable its miniaturization.
    https://www.tindie.com/products/KnivD/configum-fpga-module/
    Very nice board.
    However, IIRC the P1V will require the SLX25 as the SLX16 has only 14K LUTs. The 25K LUT Cyclone IV only fits P2 with 1 cog.
    So I am unsure how this helps the P1V or P2 ???

    PS You should start a new thread in Propeller 1 as that may be more useful to others. The Propeller 2 is a very specific thread with fewer readers.
  • cgraceycgracey Posts: 14,155
    New version 12 is posted at top of thread.

    Tomorrow, I will post a little program which programs a SPI flash with a blinker program, so you can see the SPI booter function in action.
  • jmgjmg Posts: 15,173
    edited 2016-10-03 04:25
    cgracey wrote: »
    New version 12 is posted at top of thread.

    Tomorrow, I will post a little program which programs a SPI flash with a blinker program, so you can see the SPI booter function in action.
    Nice milestone.

    Where is the info on the half-duplex serial handshakes/tokens, and the timings for turnaround.
    I'll update my rough One-Pin Serial example, so it is P2-compliant.
    - found it in docs.google link in #1..

  • cgraceycgracey Posts: 14,155
    jmg wrote: »
    cgracey wrote: »
    New version 12 is posted at top of thread.

    Tomorrow, I will post a little program which programs a SPI flash with a blinker program, so you can see the SPI booter function in action.
    Nice milestone.

    Where is the info on the half-duplex serial handshakes/tokens, and the timings for turnaround.
    I'll update my rough One-Pin Serial example, so it is P2-compliant.
    - found it in docs.google link in #1..

    It's in the Google doc, at the end. You just put a "-" between the command word and the first hex value. The booter waits 10ms, in any case, before responding serially.
Sign In or Register to comment.