Shop OBEX P1 Docs P2 Docs Learn Events
Tachyon NEON V5 (FAT32 and Ethernet Servers in 32kB EEPROM!) — Parallax Forums

Tachyon NEON V5 (FAT32 and Ethernet Servers in 32kB EEPROM!)

Peter JakackiPeter Jakacki Posts: 10,193
edited 2018-03-19 03:53 in Propeller 1
Tachyon NEON

V4 introduced wordcode vs bytecode which has proved to be faster and more compact overall but the thing that I always wanted to improve upon too was the memory hogging dictionary. While I have a COMPACT utility to transfer the existing dictionary over to EEPROM and use a hash table to index candidates it is still a hybrid dictionary with newer entries saved in hub RAM. Now it's the hub RAM that needs to be utilized to the max for code execution and buffers so it would be nice if there was a standardized way of accessing the dictionary so that it could be in any memory space. This is the conclusion I've come to and what I am proposing amongst other enhancements for Tachyon NEON V5.

Dictionary entries are fixed length and encoded with a standard record of 8 bytes that holds 5 characters or an extended record of 16 bytes that holds up to 15 characters. Compare this with the minimum variable record in V4 which requires 6 bytes for 1 character and 10 bytes for 5 characters. So V5 would mostly compare a long with a long rather than byte by byte comparisons. The method allows for optional hash table lookup in ram or elsewhere too for faster compile times. Also there will be no unique headers, if a new definition is created with an existing name then that header will simply be reused which is also good for temporary or local names.

Part of the dictionary overall also involves adding a vocabulary stack so that we can have multiple vocabularies over multiple memories and specify which ones to use and in which order.

I did an analysis of header sizes and found that most of them are 1 to 5 characters in length so it seems practical to encode up to 5 characters into a single long and forego the count byte. This leaves one byte for the attributes and 3 bytes for the wordcode.

Number of characters in kernel names
1=33   ****************
 2=120  ************************************************************
 3=138  *********************************************************************
 4=139  *********************************************************************
 5=152  ****************************************************************************
 6=85   ******************************************
 7=77   **************************************
 8=26   *************
 9=7    ***
10=2    *

Wait on, a wordcode is only 16 bits, what's the other byte for? That's for the extended addressing where code can reside in eeprom up to 1MB and the rest in SD or serial flash up to 16MB. This code will be a little slower to execute but once cached in a sector buffer could run almost as fast as hub code but we will have to see how it goes. The EXIT routine will detect 24-bit addresses and use a procedure to return to extended code and other methods are also used. Code and execution speed in hub space will be practically unaffected by this.

However the main thrust of V5 is to have an absolutely minimal dictionary footprint in hub memory if desired although this does not affect a basic Tachyon system, it simply allows for the system to be expanded easily. Even if the dictionary was moved to EEPROM after EXTEND you could still have very fast searching and compiling as the block mode could create a hash table first to select only a handful of candidates from slow EEPROM for comparison which would only need 4 bytes read in at a time initially at around 200us each. The biggest problem with EEPROM access is the setup to select a device, address the location, and then reselect in read mode, this needs to be avoided as much as possible.

<More information>




«13456725

Comments

  • ErNaErNa Posts: 1,738
    I hope to gain a more simple entry into Tachyon by mapping the concepts to a concept map. In a first step, this map: http://cmapspublic3.ihmc.us/rid=1RMDJ2BPZ-1S1RHW1-1Q0K/Tachyon_NEON_Entry.cmap only reflects the text, Peter wrote and I step by step will try to show the context ot the different statements.
  • Cluso99Cluso99 Posts: 18,066
    Wow Peter. Another nice improvement:)
  • jmgjmg Posts: 15,140
    edited 2017-08-16 22:00

    Wait on, a wordcode is only 16 bits, what's the other byte for? That's for the extended addressing where code can reside in eeprom up to 1MB and the rest in SD or serial flash up to 16MB. This code will be a little slower to execute but once cached in a sector buffer could run almost as fast as hub code but we will have to see how it goes. The EXIT routine will detect 24-bit addresses and use a procedure to return to extended code and other methods are also used. Code and execution speed in hub space will be practically unaffected by this.
    That sounds very good, and with QuadSPI and Dual Edge variants, as well as HyperFLASH at 8b transfers, the time to cache could be made quiet small.
    Does it cache just as much as it needs to ?

    i2c memory seems to be slow hitting the next 3.4MB/s specs - has anyone tried over-clocking common i2c parts above their 1M spec ?
    I can find FRAM & nvSRAM parts spec'd to 3.4MHz, so they could be used for testing, but they are not what could be called cheap...

  • jmg wrote: »
    That sounds very good, and with QuadSPI and Dual Edge variants, as well as HyperFLASH at 8b transfers, the time to cache could be made quiet small.
    Does it cache just as much as it needs to ?

    While Tachyon will use I2C EEPROM as it's primary "secondary memory" it is not because it is fast or cheap, but simply because every Prop system has it and the chip can be easily substituted with a higher capacity etc. But I do have QSPI Flash on many boards although they only run on 1-bit mode but they would be a lot faster than I2C already. However I will at some point try out QSPI mode after I get this up and running.
  • jmgjmg Posts: 15,140
    edited 2017-08-17 04:33
    ... However I will at some point try out QSPI mode after I get this up and running.
    DTR parts should be slightly faster in SW access...

    I see Winbond have some DTR flash parts, but the part-codes seem not easy to find.
    As best I can decode, the final letter is M on DTR+Quad parts and Q on non-DTR (just Quad)
    Chipstop show
    W25Q16JVSSIM (SOIC-8 208-mil) 33c/100+
    W25Q16JVZPIM ZP WSON-8 6x5-mm 16M-bit $0.677/10+

    and then there are new Octoflash parts :

    http://www.macronix.com/en-us/products/NOR-Flash/Pages/OctaFlash.aspx#3V
    which shows 512Mb (prodn) and 256Mb, 128Mb planned.
    These parts DO have a 300mil 16-SOP package choice, which is easier to handle and prototype with than 6x8mm 24-TFBGA(5x5)

    These can support 1b SPI, mode, for backward compatible use, and they also have the 8b Dual-Edge modes, for highest throughput.


  • Started making some progress on V5 mainly with the dictionary. I settled on names truncated to 8 characters for now in a 12-byte fixed record and this seems to work really well. I already have EXTEND and EASYFILE running nicely. If I really want to I can encode 12 character names such as "LONGER_NAMES" into the same header but I don't think I really need too.

    HELP is designed to be rather general and at the moment it lists all the words of the dictionary in a more useful format too. Here's a small sample:
    $0062D4: FALSE      h.......  == $0000 (0)
    $0062E0: TRUE       h....... PASM $0CD
    $0062EC: ..         h.i.....   $0011A4
    $0062F8: s          h.......   $000304
    $006304: |<         h....... PASM $099
    $006310: ]          h.i.....   $001BFE
    $00631C: =[         h.i.....   $001BF4
    $006328: BREAK      h.i.....   $001BFE
    $006334: CASE       h.i.....   $001BF4
    $006340: SWITCH@    h.......   $001BEE
    $00634C: SWITCH     h.......   $001BE8
    $0063F4: NFA>ATR    h.......   $001B0C
    $006400: +NFA       h.......   $001B06
    $00640C: CLKMHZ     h.......  == $0060 (96)
    $006418: CLKFREQ    h.......   $001AFE
    

    Internally the search methods allow for external memory methods and code execution is being upgraded to 32-bit addresses as well. My next step is to selectively assign some words to the EEPROM dictionary and others to RAM and perhaps even SD to test dictionary handlers. BTW, the compilation speed is very fast, I can set the line delay to 5ms using GtkTerm at 115200 baud which will be the same for TeraTerm so that it takes 9 seconds to download and compile EASYFILE.
      Propeller .:.:--TACHYON--:.:. Forth V5.0 NEON 500170820.1200
    Loading cog 3 E50A F32     
    Mounted D20D.AA60-CB4A.74CF mkfs.fat IOT5500     FAT32   7,947MB (4,096/cluster)
    *** MODULES ***
                           Propeller .:.:--TACHYON--:.:. Forth V5.0 NEON 500170820.1200
    3778: FILE.fth        SDHC card + FAT32 Virtual Memory Access File System Layer V1.2 170807-1600 
    1AC0: EXTE.fth        Primary extensions to TACHYON V4.5 kernel  - 170814-1100
    
    AUTORUN BOOT
    FREQ = 96.0MHz
    *** INITS ***
    INIT#0 3CFA MOUNT
    *** ROMS ***
    E01C: VGA32x15    848
    E37C: HSUART      392
    E514: F32        1900
    *** I2C ***
    $A0 EEPROM
    I/O =  31:UHU~ 27:HHLU 23:~~~U 19:~~UU 15:~~~~ 11:~~~~ 7:~~~~ 3:~~UU
    INTERCOM: &00.00.00.00 @2,000,000
    
    CODE:$483A = 17978 bytes 
    NAME:$4CB4 = 10060 bytes 
    DATA:$7AD2 = 1218 bytes 
    FREE:      = 1146 bytes 
     Data Stack (0)
    2001/01/01 MON 00:00:00
    --------------------------------------------------------------------------------
    ..
    
  • MJBMJB Posts: 1,235
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...
  • MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Ah yes, but I thought I would open up a lot more memory for the Internet of "EveryThing" :)
    Even with a flexible dictionary structure I can do a lot more as well after which I can explore my VMM (Virtual Memory Model).
  • MJBMJB Posts: 1,235
    MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Ah yes, but I thought I would open up a lot more memory for the Internet of "EveryThing" :)
    Even with a flexible dictionary structure I can do a lot more as well after which I can explore my VMM (Virtual Memory Model).
    I had the scriptable webserver running on Tachyon 2.3 with Spinneret - then 'everything' changed and was in constant flow.
    And since this integrates into EASYNET using some internal SD accelleration and does not sit on top,
    it really makes sense to get it running and then keep it as part of EASYNET, so any future system change can be handled by you @Peter.
    It is not a lot of code actually. When the hook is in the html file reader everything else is handled in the script parts inside the HTML files.
    So partially dynamic HTML or JS files can easily be scripted using Tachyon as the scripting language. JSON can be used as well. A JSON writer for Tachyon exists.
  • MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Yes this would be great to include in easynet even conditionally. It seems V5 will have plenty of space for all sorts of add-ons. I'm interested in this doc as well.




  • MJBMJB Posts: 1,235
    D.P wrote: »
    MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Yes this would be great to include in easynet even conditionally. It seems V5 will have plenty of space for all sorts of add-ons. I'm interested in this doc as well.




    this is a really tiny extension to the code
  • MJB wrote: »
    D.P wrote: »
    MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Yes this would be great to include in easynet even conditionally. It seems V5 will have plenty of space for all sorts of add-ons. I'm interested in this doc as well.

    this is a really tiny extension to the code

    I will take a look at it today on V4.5.
  • MJBMJB Posts: 1,235
    edited 2017-08-21 00:01
    MJB wrote: »
    D.P wrote: »
    MJB wrote: »
    IF you had a little time we could also have some progress on the scriptable web server ...
    I shared the google doc with you ...

    Yes this would be great to include in easynet even conditionally. It seems V5 will have plenty of space for all sorts of add-ons. I'm interested in this doc as well.

    this is a really tiny extension to the code

    I will take a look at it today on V4.5.
    I will be around from 8:00 CEDT = 15:00 Melbourne ??
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-02-26 05:28
    While I have been working on V4.8 with a lot of changes to the dictionary to save memory I see that there are too many changes and that it is a good candidate for a V5 instead. So I have promoted this version to V5.1 bypassing the V5.0 stuff I was looking at last year.

    The dictionary searching is much faster and is the same even when pasting code outside of a TACHYON END block load since V4 load a fast search module only during block mode. The headers in the dictionary now combine the name count, the attributes, and a vocab ID. Standard Forth colon definitions are either pub, pri, pre, or module to set attributes for public, private, preemptive, or to identify a module header which used to be anything that ended in ".fth". The module attribute will also control the scope for word searches so that the module can be searched first, and if not found then the whole dictionary.

    There's still lots of optimization I intend to do but it is fully functional and tested so far. Why would you use this version? Various words have been sped up, it loads faster, and includes extras while having more memory left over, even before RECLAIM'ing. I noticed SPIPINS that sets up the masks for the SPI instructions etc used to take over 80us but now it executes in less than 20us. Okay, many are still the same speed, but I've been tidying up some of the tardy ones.

    NEON-1.png

    Neon colors in the print? Nice and easy (figure it out):
    byte ncol
    pub NEON    EMIT: [CON ncol C@ 3 AND " 6325" + C@ >N PEN EMIT ncol C++ CON]  ;
    

    1560 x 1052 - 43K
  • Cluso99Cluso99 Posts: 18,066
    What? Neon?

    Nothing to do with space probes :(
  • I have a board with a 6.25MHz crystal -- Tachyon shows the speed as 00.00MHz. Small thing I know, but begs the question: Is there an issue running Tachyon at 100MHz?
  • JonnyMac wrote: »
    I have a board with a 6.25MHz crystal -- Tachyon shows the speed as 00.00MHz. Small thing I know, but begs the question: Is there an issue running Tachyon at 100MHz?

    More like oversight I would say :) I formatted the report to always use 2+2 digits for the frequency!
    CLKFREQ 10,000 /  .AS" ##.##MHZ"
    

    I will just change the .AS" format so that it reports correctly as in this test:
    ...  100.00 .AS" *|#.##MHZ"   100.00MHZ ok
    ...  96.00 .AS" *|#.##MHZ"   96.00MHZ ok
    ...  80.00 .AS" *|#.##MHZ"   80.00MHZ ok
    
  • MJBMJB Posts: 1,235
    edited 2018-03-16 00:58
    just had a look at latest V5.1 spin

    there seems to be global replace error starting with your name and then many places ...
    2012..2017 Peter Jalphaacki
    
    I did a compare with 5.1 from Feb.27
  • MJBMJB Posts: 1,235
    finally Tachyon is grown up and got a
    CREATE ... DOES>  ....
    

    great
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-03-16 04:57
    @mjb - yep, I noticed something was wrong :) Changed all the "ak" keytable words to "alphas" but must have missed the "whole word" setting in Atom editor. btw, I'm slowly building up a specific Tachyon grammar for highlighting.

    I started out with MAKE: and DOES: but by renaming the kernel CREATE to CREATE: I was able to use the more familiar CREATE DOES> words. The only difference is that the DOES> portion is called by any function created with CREATE but does not have a "parameter field address" on the stack due to the way that Tachyon functions. It is easy enough though to do a simple R> to get that parameter field address which is usually all that is required as in the case of TABLE etc.

    There's also a simple decompiler built into EXTEND that uses <100 bytes but is still quite effective and includes ANSI color coding. Trouble is, it shows you just how simple some of these functions are:
    ...  DECOMP DOES>
    1B4E: R>
    1B50: 0
    1B52: REG
    1B54: W@
    1B56: W!
    1B58: EXIT
    

    WWORDS also gives a color coded formatted listing of the dictionary.
  • MJBMJB Posts: 1,235
    edited 2018-03-16 13:21
    ... btw, I'm slowly building up a specific Tachyon grammar for highlighting.
    I am using Notepad++ and made a syntax highlighter and it also provides a nice function/word list in a side window for direct access

    will investigate CREATE DOES> in detail next week
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-03-18 04:00
    NEON has enhanced the simple DECOMP and incorporated it into the HELP word so that HELP will list the word the same way WWORDS does plus performs a DECOMP on that word. With a little extra I may add a compact stack hint for most words as well.

    Most of the development/debug tools are in the TOOLS module at the end of EXTEND so they may be forgotten without impacting code to save a couple of kB of memory in addition to the 1.4kB that is saved by RECLAIMing private words. Also the VGA, ANYKEY, and LCD modules are placed at the end of EXTEND before TOOLS so these may be safely forgotten as well. If you do all this it will provide 11.3kB free vs the 6.2kB free with all the modules etc but of course there is no fancy report on boot, just the version and memory stats.

    Remember too that the top 3kB from $7400 onwards is dedicated to variables and various buffers etc.

    Full version of EXTEND
    CODE:$4012 = 15890 bytes 
    NAME:$5830 = 7120 bytes 
    DATA:$7687 = 375 bytes 
    FREE:      = 6174 bytes 
    

    Minimum version of EXTEND
    CODE:$32CA = 12490 bytes 
    NAME:$5F02 = 5374 bytes 
    DATA:$7687 = 375 bytes 
    FREE:      = 11320 bytes 
    

    HELP word (still adding stack hints)
    HELP EVAL$
    5D40 2FCE 0 EVAL$ ( str -- )
    2FCE: DUP
    2FD0: FROM
    2FD2: LEN$
    2FD4: FOR
    2FD6: I
    2FD8: C@
    2FDA: doKEY
    2FDC: NEXT
    2FDE: $0020
    2FE0: doKEY
    2FE2: 13
    2FE4: doKEY EXIT
    
    NEON-180318.png
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-03-19 00:11
    Just to see how much memory would be free after releasing memory by forgetting the reporting tools and reclaiming private words, and then loading EASYFILE FAT32 file system I still end up with over 4kB of memory (over 6kB if I forget the VGA/LCD/KEY modules). I could further compact the dictionary into EEPROM which would yield over 6kB extra leaving over 10k free. That's a lot packed into the itty-bitty P1 memory and a lot left over!
      Propeller .:.:--TACHYON--:.:. Forth V5.1 NEON 510180318.1300
    CODE:$4A90 = 18576 bytes 
    NAME:$5AF6 = 6410 bytes 
    DATA:$78CD = 957 bytes 
    FREE:      = 4198 bytes 
     Data Stack (0)
    
    CARD: SU32G SD03.80 #2C59.C01F 2013/4 !C0C3.4040 2,308us
    FAT: #6332.3935          PBJTAB1     FAT32   3191078912.0 bytes (32kB clusters)
    
    --------------------------------------------------------------------------------
    ...  WORDS 
     DICTIONARY WORDS @5AF6 
     QV FL SAVETEXT FRUN .FILES .FX .FILE lss VOLNAME! ls .LIST .FNAME (SLIST) DIR cd$ cd pwd cat (cat) FPRINT$ FLOAD FCOPY FCOPY$ RENAME RENAME$ 
    APPEND -FERASE FMAKE$ FOPEN FOPEN$ mk RW RO FOPEN# FCREATE$ MAKE FCLOSE >FILE FILE> FPUTB FPUT FGET FREM fwrite fread FSTAMP DIR? FSIZE! Updat
    eDir OpenDir FILE$ ?MOUNT MOUNT ROOT FMAX@ CLUST>SECT @FAT @BOOT @ROOT FS FS@ FSW@ FSC@ FS! FSC! FSADR SD XW@ XC! XC@ X! X@ XADR FSECT@ FSIZE@
     FSIZE dirbuf dirfsa @FILE FILE# FILE SDRD SECTOR SCAN! FLUSH WRSECT SDWR !SD ACMD CMD SDPINS SDIO32 SDIO8 CARD? sdbuf _sdpins ocr csd cid #fi
    les BLKSIZ OVER+ @rest EASYFILE !LCD LCD LCDBL CHARLCD KEYPAD KEYPAD@ !KEYPAD KEYMUTE KEYBEEP BEEPER ANYKEYS !VGA VGA VCHAR PLOT VXY HUE rows 
    cols screen VGAPIN @VP VGATEXT BOOT DEFER FSQRT FSIN F/ F* F- F+ F> >F FCMD LOADCOGS LOADCOG PWM! PWM PWM.START PWM% OK EVAL$ tib LEDS LED LED
    PIN .DHT DHT DISTANCE PING ID! INTERCOM! +NFA CTYPE PRINT$ MLEN$ CPA MU@ MW@ MC@ MC@++ +POLL ?POLL TIMERJOB !RTC .TEMP 'F 'C STAMP@ .DT TZ tz$
     .ASMONTH .DATE .TIME .DAY DAY DAY@ SUN SAT FRI THU WED TUE MON DATE! SDT! DT! TIME! DATE@ HMS TIME@ DEC>BCD BCD>DEC RTC RTC@ RTC! WRRTC RDRTC
     NORTC RTCEE MCP79410 DS3231 @rtc runtime WATCHDOG TIMEOUT? ALARM COUNTUP TIMEOUT timers TIMER AUTORUN IO! IO@ EE BACKUP ?BACKUP CONBAUD ECOPY
     EFILL ELOAD ESAVE ep I2CPINS EEPROM ESAVEB E! E@ EW@ EW! EC@ EC! EERD @EEWAIT @EE eeadr I2C400 I2C100 I2CFAST nakI2C@ I2C@ ackI2C@ I2C! I2C!?
     I2C> <I2C> <I2C ?I2C i2cflg MODPINS SETPINS SPIPINS @SPISCK @SCL @CNT @CE @MISO @MOSI @SCK *SDA *SCL BLINK RINGS RING SIREN WARBLE BIP BEEPS 
    TONE BEEP CLICK SPKR MUTE HZ KHZ MHZ VOLTS FRQ DAC! BPIN APIN PLLDIV PLL DUTY DIFF CTRMODE NCO B A SERIN ISERIAL SERIAL ISEROUT SEROUT SERBAUD
     baudcnt +VECTOR REVECTOR (FORGET) FORGET RECLAIM STRIP INIT -INIT +INIT .CFA .HEAD$ ?ANSI MARGINS WRAP BOLD REVERSE PLAIN CURSOR ERLINE ERSCN
     CLS XY PAPER PEN HOME ESC white cyan magenta blue yellow green red black .BIN32 .BIN16 PRINT& .DEC2. .DEC2 .DECL .DEC4 .AS" .AS COGINIT TASKR
    EGS RUN: RUN TASK? RND PINS@ OUT PIN! CNT? =CNT CNT@ P! P@ VSCL VCFG PHSB PHSA FRQB FRQA CTRB CTRA DIRB DIRA OUTB OUTA INB INA CNT PAR SPR LEF
    T$ RIGHT$ MID$ +CHAR APPEND$ $= COMPARE$ LOCATE$ $! COPY$ NULL$ CON] CON [CON boot NULLOUT EMIT: KEY: M MB KB 1M us R@ INVERT MOD 2OVER 3DUP 2
    SWAP @. => <= AVG C~~ W~~ ~~ W>L B>L B>W W>B L>W >W 3++ 3@ 3! @3 2++ 2@ 2! @2 C-- C++ W-- W++ -- ++ A> >A A@ FALSE TRUE |< ] =[ SWITCHES CASES
     BREAK CASE SWITCH@ SWITCH STRING TABLE |<= WRAP DS DS+ clong cword cbytes CARRAY CREATE DOES> QUIET NUM>STR NUMBER .WORD: .WORDSP LIMIT >| WW
    ! D@ D! U@ U! CLKMHZ CLKFREQ ANYCASE PUBLIC PRIVATE EXTEND " ." PRINT" == vars org @org byte word long bytes words longs , || | C, W, L, ALIAS
     : pre pub pri module ; [C] ' ['] NFA' NFA$ NOP DUP 2DUP OVER DROP 2DROP SWAP ROT -ROT NIP 3DROP ?DUP 3RD 4TH >R R> !RP !SP AND ANDN OR XOR RO
    L ROR SHR >> 8>> SHL << 8<< 2/ 2* 4* SAR SPLIT9 REV MASK >N >B 0= NOT 1+ 1- + - 2+ 4+ 2- * UM* U/ U/MOD / */ UM*/ IF ELSE THEN ENDIF BEGIN UNT
    IL AGAIN WHILE REPEAT FROM BY FOR NEXT LOOP I+ I J K LEAVE FOR@ FOR! BY! LP! IC! IC@ C@ W@ @ C+! C! C@++ W+! W! +! ! TX! 1! 1@ 1++ C~ W~ ~ BIT
    ? SET? SET CLR TOG BIT! CMOVE <CMOVE ERASE FILL RESET 0EXIT EXIT ?EXIT CALL JUMP ?JUMP WAITLO WAITHI RUNMOD (EMIT) (EMITX) LOADMOD COG@ COG! C
    OGSTOP pCOGINIT COGID REBOOT CLK CLKSET DELTA WAITCNT LAP LAP@ CLOCK CLKIN CLKOUT H L P T F SHROUT SHRINP OUTSET OUTCLR OUTPUTS INPUTS HIGH LO
    W FLOAT PIN@ IN SPIWRB SPIWR16 SPIWR32 SPIWR SPIRD SPICE 0<> <> WITHIN U> = > < U< 0< EMIT SPACE SPACES CR KEY WKEY (KEY) KEY! doKEY CLS EMITS
     DOT BELL <CR> TAB TABS XTAB SPINNER .HEX .BYTE .WORD .LONG .WORD$ @PAD HOLD >CHAR #> <# # #S <D> PRINT$ CTYPE .NFA LEN$ U. .DP PRINT . CREATE
    : CREATE$ GETWORD SEARCH CFA CPA +NFA NFA ALIGNORG ALIGN @NAMES names WORDS ALLOT ALLOCATED HERE [W] [B] ["] KEY$ [W,] VER .VER TACHYON END CO
    NSOLE ECHO ms s DISCARD TASK REG BUFFERS COLD HEX DECIMAL BINARY RAM DUMP: DUMPX DUMP DUMPW DUMPL DUMPC DUMPA DUMPAW QD DEPTH .S DEBUG .STATS 
    IDLE NOOP --- \ '' ( { } IFDEF IFNDEF OFF ON GRAB ABS ?NEGATE NEGATE MIN MAX MINS MAXS ADDABS SUBABS CMPSUB UM/MOD64 UM/MOD32 WAITVID [WS2812]
     [SDRDF] [SDRD] [SDWR] [SDIO] [SSD!] [PWM32] [PWM32!] [PLOT] [ROL3] [CAP] [WAV] [MCP32] [RCTIME] errors uhere uthere flags prompt uauto keypol
    l lastkey rxpars rx id keytable @WORD uemit ukey base num uswitch undef (.") (") ... V5 *end* 
     ok
    ...
    
  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-03-19 03:36
    For the very first time I think it has become possible to fit a complete Tachyon system with EASYFILE and EASYNET Ethernet servers into the lower 32kB of EEPROM without requiring upper EEPROM. Previously the dictionary had to be compacted into upper EEPROM before EASYFILE could be compiled. Since the SD card and WIZnet pins are configurable at runtime this means that I can create a standard binary image that can be loaded easily with Spin tool, all set , ready to go!

    What's more the server protocols are now independent modules which can be loaded on top of EASYNET, no need to have it compiled at the same time. This means that a protocol can be enhanced or added to suit user requirements.

    By loading and running the COMPACT extension a further 7kB of dictionary can be moved into upper EEPROM leaving 8kB free.

    After I play with this a bit more I will make and upload an image into the V5 directory.
    Propeller .:.:--TACHYON--:.:. Forth V5.1 NEON 510180318.1300
    CODE:$5424 = 21028 bytes 
    NAME:$5824 = 7132 bytes 
    DATA:$799A = 1162 bytes 
    FREE:      = 1024 bytes 
     Data Stack (0)
    
    CARD: SU32G SD03.80 #2C59.C01F 2013/4 !C0C3.4040 2,310us
    FAT: #6332.3935          PBJTAB1     FAT32   31,910,789,120 bytes (32kB clusters)
    
    --------------------------------------------------------------------------------
    ...  ifconfig 
    NETWORK STATUS:
    LINK DOWN
    HARDWARE: WIZnet W5500 V0
    SRC IP    &0.0.0.0
    MASK      &0.0.0.0
    GATEWAY   &0.0.0.0
    MAC       00.00.00.00.00.00.
    SKT HH:MM:SS MODE  PORT  DEST TXRD TXWR RXRD RXWR RXSZ  IR STATUS            IP ADDR
     ok
    ...  WORDS 
     DICTIONARY WORDS @5824 
     !FTP STOR RETR SIZE LIST ?DISC PWD CWD CDUP MDTM FEAT SYST PASV PORT TYPE PASS USER EASYNET !EASYNET RESTART ?EASYNET constat CONNECT DISC? B
    YE CONNECTED? KEEPALIVE LANSKT BLKSEND SENDFILE EASYNET HTTP TELNET FTPDAT FTP NETMAN .SOCKETS ifconfig .SKT .SOCKET .SKTHD RXSKT WIZ !WIZ !WI
    ZIP !WIZIO WCOLD LANCON LANCONKEY LANCONEMIT LAN LANKEY LANEMIT LANSEND ?SEND ?SENDPOLL WAITSEND autosend !TXWR @txwr SetPORT PORT! sRECV sSEN
    DKEEP sSENDMAC sSEND sCLOSE sDISCON sCONNECT sLISTEN sOPEN UDP TCP @SOCKET SKT@ SKT SOCKET SIP MAC SUBNET GATEWAY wMODE @subnet @sip @gateway 
    LWRITE LREAD L@ LW@ LC@ L! LW! LC! @RX @TX @SKT @COMMON WIZPINS LANLED LANLED.ON RDYLED W5500 QV FL SAVETEXT FRUN .FILES .FX .FILE lss VOLNAME
    ! ls .LIST (SLIST) DIR .FNAME cd$ cd pwd cat (cat) FPRINT$ FLOAD FCOPY FCOPY$ RENAME RENAME$ APPEND -FERASE FMAKE$ FOPEN FOPEN$ mk RW RO FOPEN
    # FCREATE$ MAKE FCLOSE >FILE FILE> FPUTB FPUT FGET FREM fwrite fread FSTAMP DIR? FSIZE! UpdateDir OpenDir FILE$ ?MOUNT MOUNT ROOT FMAX@ CLUST>
    SECT @FAT @BOOT @ROOT FS FS@ FSW@ FSC@ FS! FSC! FSADR SD XW@ XC! XC@ X! X@ XADR FSECT@ FSIZE@ FSIZE dirbuf dirfsa @FILE FILE# FILE SECTOR SCAN
    ! FLUSH WRSECT SDWR !SD ACMD CMD SDPINS SDIO32 SDIO8 CARD? sdbuf _sdpins ocr csd cid #files BLKSIZ OVER+ @rest EASYFILE BOOT DEFER FSQRT FSIN 
    F/ F* F- F+ F> >F FCMD LOADCOGS LOADCOG PWM! PWM PWM.START PWM% OK EVAL$ tib LEDS LED LEDPIN .DHT DHT DISTANCE PING ID! INTERCOM! +NFA CTYPE P
    RINT$ MLEN$ CPA MU@ MW@ MC@ MC@++ +POLL ?POLL TIMERJOB !RTC .TEMP 'F 'C STAMP@ .DT TZ tz$ .ASMONTH .DATE .TIME .DAY DAY DAY@ SUN SAT FRI THU W
    ED TUE MON DATE! SDT! DT! TIME! DATE@ HMS TIME@ DEC>BCD BCD>DEC RTC RTC@ RTC! WRRTC RDRTC NORTC RTCEE MCP79410 DS3231 @rtc runtime WATCHDOG TI
    MEOUT? ALARM COUNTUP TIMEOUT timers TIMER AUTORUN IO! IO@ EE BACKUP ?BACKUP CONBAUD ECOPY EFILL ELOAD ESAVE ep I2CPINS EEPROM ESAVEB E! E@ EW@
     EW! EC@ EC! EERD @EEWAIT @EE eeadr I2C400 I2C100 I2CFAST nakI2C@ I2C@ ackI2C@ I2C! I2C!? I2C> <I2C> <I2C ?I2C i2cflg MODPINS SETPINS SPIPINS 
    @SPISCK @SCL @CNT @CE @MISO @MOSI @SCK *SDA *SCL BLINK RINGS RING SIREN WARBLE BIP BEEPS TONE BEEP CLICK SPKR MUTE HZ KHZ MHZ VOLTS FRQ DAC! B
    PIN APIN PLLDIV PLL DUTY DIFF CTRMODE NCO B A SERIN ISERIAL SERIAL ISEROUT SEROUT SERBAUD baudcnt +VECTOR REVECTOR (FORGET) FORGET RECLAIM STR
    IP INIT -INIT +INIT .CFA .HEAD$ ?ANSI MARGINS WRAP BOLD REVERSE PLAIN CURSOR ERLINE ERSCN CLS XY PAPER PEN HOME ESC white cyan magenta blue ye
    llow green red black .BIN32 .BIN16 PRINT& .DEC2. .DEC2 .DECL .DEC4 .AS" .AS COGINIT TASKREGS RUN: RUN TASK? RND PINS@ OUT PIN! CNT? =CNT CNT@ 
    P! P@ VSCL VCFG PHSB PHSA FRQB FRQA CTRB CTRA DIRB DIRA OUTB OUTA INB INA CNT PAR SPR LEFT$ RIGHT$ MID$ +CHAR APPEND$ $= COMPARE$ LOCATE$ $! C
    OPY$ NULL$ CON] CON [CON boot NULLOUT EMIT: KEY: M MB KB 1M us R@ INVERT MOD 2OVER 3DUP 2SWAP @. => <= AVG C~~ W~~ ~~ W>L B>L B>W W>B L>W >W 3
    ++ 3@ 3! @3 2++ 2@ 2! @2 C-- C++ W-- W++ -- ++ A> >A A@ FALSE TRUE |< ] =[ SWITCHES CASES BREAK CASE SWITCH@ SWITCH STRING TABLE |<= WRAP DS D
    S+ clong cword cbytes CARRAY CREATE DOES> SHORTCUT QUIET NUM>STR NUMBER .WORD: .WORDSP LIMIT >| WW! D@ D! U@ U! CLKMHZ CLKFREQ ANYCASE PUBLIC 
    PRIVATE EXTEND " ." PRINT" == vars org @org byte word long bytes words longs , || | C, W, L, ALIAS : pre pub pri module ; [C] ' ['] NFA' NFA$ 
    NOP DUP 2DUP OVER DROP 2DROP SWAP ROT -ROT NIP 3DROP ?DUP 3RD 4TH >R R> !RP !SP AND ANDN OR XOR ROL ROR SHR >> 8>> SHL << 8<< 2/ 2* 4* SAR SPL
    IT9 REV MASK >N >B 0= NOT 1+ 1- + - 2+ 4+ 2- * UM* U/ U/MOD / */ UM*/ IF ELSE THEN ENDIF BEGIN UNTIL AGAIN WHILE REPEAT FROM BY FOR NEXT LOOP 
    I+ I J K LEAVE FOR@ FOR! BY! LP! IC! IC@ C@ W@ @ C+! C! C@++ W+! W! +! ! TX! 1! 1@ 1++ C~ W~ ~ BIT? SET? SET CLR TOG BIT! CMOVE <CMOVE ERASE F
    ILL RESET 0EXIT EXIT ?EXIT CALL JUMP ?JUMP WAITLO WAITHI RUNMOD (EMIT) (EMITX) LOADMOD COG@ COG! COGSTOP pCOGINIT COGID REBOOT CLK CLKSET DELT
    A WAITCNT LAP LAP@ CLOCK CLKIN CLKOUT H L P T F SHROUT SHRINP OUTSET OUTCLR OUTPUTS INPUTS HIGH LOW FLOAT PIN@ IN SPIWRB SPIWR16 SPIWR32 SPIWR
     SPIRD SPICE 0<> <> WITHIN U> = > < U< 0< EMIT SPACE SPACES CR KEY WKEY (KEY) KEY! doKEY CLS EMITS DOT BELL <CR> TAB TABS XTAB SPINNER .HEX .B
    YTE .WORD .LONG .WORD$ @PAD HOLD >CHAR #> <# # #S <D> PRINT$ CTYPE .NFA LEN$ U. .DP PRINT . CREATE: CREATE$ GETWORD SEARCH CFA CPA +NFA NFA AL
    IGNORG ALIGN @NAMES names WORDS ALLOT ALLOCATED HERE [W] [B] ["] KEY$ [W,] VER .VER TACHYON END CONSOLE ECHO ms s DISCARD TASK REG BUFFERS COL
    D HEX DECIMAL BINARY RAM DUMP: DUMPX DUMP DUMPW DUMPL DUMPC DUMPA DUMPAW QD DEPTH .S DEBUG .STATS IDLE NOOP --- \ '' ( { } IFDEF IFNDEF OFF ON
     GRAB ABS ?NEGATE NEGATE MIN MAX MINS MAXS ADDABS SUBABS CMPSUB UM/MOD64 UM/MOD32 WAITVID [WS2812] [SDRDF] [SDRD] [SDWR] [SDIO] [SSD!] [PWM32]
     [PWM32!] [PLOT] [ROL3] [CAP] [WAV] [MCP32] [RCTIME] errors uhere uthere flags prompt uauto keypoll lastkey rxpars rx id keytable @WORD uemit 
    ukey base num uswitch undef (.") (") ... V5 *end* 
     ok
    ...
    
  • Cluso99Cluso99 Posts: 18,066
    Wow Peter.

    By V9 you will not need any EEPROM at all ! ;););)
  • Cluso99 wrote: »
    Wow Peter.

    By V9 you will not need any EEPROM at all ! ;););)

    Yes, and imagine where would we be if we had P1B with extra memory and I/O all those years ago when P2 "was almost here"? :)
    I know there would have been a lot more still with us who have since fallen by the wayside, and a lot more newer ones, which would have created a perfect market for that future day when we have P2!


  • Cluso99Cluso99 Posts: 18,066
    Cluso99 wrote: »
    Wow Peter.

    By V9 you will not need any EEPROM at all ! ;););)

    Yes, and imagine where would we be if we had P1B with extra memory and I/O all those years ago when P2 "was almost here"? :)
    I know there would have been a lot more still with us who have since fallen by the wayside, and a lot more newer ones, which would have created a perfect market for that future day when we have P2!

    Remember years ago when we were asked if we wanted a P1B first, or the P2 which was only a short time off?
    Who said go for P2??? <ducks for cover>
  • ErNaErNa Posts: 1,738
    With a next P we will enter into free trade era of Px, ... so time to wait is worth. Sometime it takes something strange to recognize, what "good" means ;-)
  • MJBMJB Posts: 1,235
    edited 2018-03-19 17:29
    For the very first time I think it has become possible to fit a complete Tachyon system with EASYFILE and EASYNET Ethernet servers into the lower 32kB of EEPROM without requiring upper EEPROM. Previously the dictionary had to be compacted into upper EEPROM before EASYFILE could be compiled. Since the SD card and WIZnet pins are configurable at runtime this means that I can create a standard binary image that can be loaded easily with Spin tool, all set , ready to go!

    What's more the server protocols are now independent modules which can be loaded on top of EASYNET, no need to have it compiled at the same time. This means that a protocol can be enhanced or added to suit user requirements.

    so it might be time soon that Tachyon is stable enough again (for a while ;-) ... )
    to get back to port the scriptable webserver extensions from Tachyon 2.7 ...

    If the HTTP/Web module is incremental it might be simpler to handle
    I will check out your code


  • Peter JakackiPeter Jakacki Posts: 10,193
    edited 2018-03-28 13:31
    @MJB - Check out V5.2 as I am using that for testing with Ethernet.

    BTW, I came across Pascal's Triangle in Forth so I converted it straight across into Tachyon and it worked first go!

    Pascal's Triangle code
    : PascTriangle
    	CR DUP FOR
    	  1 OVER 1- I - 2* SPACES I 1+
    	  FOR DUP .AS" ###`# " J I - * I 1+ / NEXT
    	  CR DROP
    	NEXT DROP
    	;
    


    Running it
    ...  16 PascTriangle
                                     1
                                   1    1
                                 1    2    1
                               1    3    3    1
                             1    4    6    4    1
                           1    5   10   10    5    1
                         1    6   15   20   15    6    1
                       1    7   21   35   35   21    7    1
                     1    8   28   56   70   56   28    8    1
                   1    9   36   84  126  126   84   36    9    1
                 1   10   45  120  210  252  210  120   45   10    1
               1   11   55  165  330  462  462  330  165   55   11    1
             1   12   66  220  495  792  924  792  495  220   66   12    1
           1   13   78  286  715 1287 1716 1716 1287  715  286   78   13    1
         1   14   91  364 1001 2002 3003 3432 3003 2002 1001  364   91   14    1
       1   15  105  455 1365 3003 5005 6435 6435 5005 3003 1365  455  105   15    1
     ok
    
Sign In or Register to comment.