Macro Assembler for P1, P2

Comments

• jmg Posts: 15,191

  2016-09-21 20:55 edited 2016-09-22 20:55

  Example of progress so far, mainly quick framework testing :
  Some web-help, and now Updated to include both IF_ Prefix and W Suffix handling.
  

  ; ~~~~~~~~~~~~~~ Now Expanded to support Suffix too.. ~~~~~~~~~~~~~~~~~
  
  ;web suggestion 
  define NR SUFF -001b 
  define WC SUFF 010b 
  define WZ SUFF 100b 
  define SUFF
  
  gParC = 1111b ; default, globalV init
  
  macro F_ZCDS? ParD*,S_Suffix*&  ; & for ParS and any suffix controls
     gZCR = 001b ; default
     match S =SUFF suffix, S_Suffix  ;splits S,
         ParS = S 
         local flags 
         iterate flag,suffix   ; comma delimited list 
             if flag >= 0 
                 gZCR = gZCR or flag 
             else  ;NR 
                 gZCR = gZCR and not -flag 
             end if 
         end iterate 
     else 
         ParS = S_Suffix 
     end match 
     gZCDS  = (gZCR shl 23) + gParC shl 18 + (ParD and 0x1ff) shl 9 + ParS and 0x1ff 
     gParC = 1111b ; restore global default, after use  
  end macro  
  
  ;           ZCRi    
  ;XOR 011011 001i cccc ddddddddd sssssssss  
  ;    109876 5432 1098 765432109 876543210  
  macro qXOR? ParD*,ParS*& ; ParC now global, ZCR via any suffix 
      F_ZCDS ParD,ParS     
      match #data, ParS  
          dd (011011_000_1b shl 22) + gZCDS  
      else    
          dd (011011_000_0b shl 22) + gZCDS  
      end match    
  end macro 
  
  ; ~~~~~~~~~ Build 32 IF_ variants ~~~~~~~~~~~~    
  iterate <Cond,wxyz>,\
  ALWAYS,1111b,NEVER,0000b,E,1010b,NE,0101b,A,0001b,B,1100b,AE,0011b,BE,1110b,C,1100b,NC,0011b,\
  Z,1010b,NZ,0101b,C_EQ_Z,1001b,C_NE_Z,0110b,C_AND_Z,1000b,C_AND_NZ,0100b,NC_AND_Z,0010b,NC_AND_NZ,0001b,\ 
  C_OR_Z,1110b,C_OR_NZ,1101b,NC_OR_Z,1011b,NC_OR_NZ,0111b,Z_EQ_C,1001b,Z_NE_C,0110b,Z_AND_C,1000b,\  
  Z_AND_NC,0010b,NZ_AND_C,0100b,NZ_AND_NC,0001b,Z_OR_C,1110b,Z_OR_NC,1011b,NZ_OR_C,1101b,NZ_OR_NC,0111b  
      macro qIF_#Cond? instruction&  
          gParC = wxyz ; apply CC value, now global
          instruction 
      end macro  
  end iterate 
  
  ; ~~~~~~~~~ Test coverage ~~~~~~~~~~~~
  qIF_Z      qXOR 0AH,#05H  WC       ;C field set    011
  qIF_Z      qXOR 0AH,#05H  WZ       ;Z field set    101 
  qIF_Z      qXOR 0AH,#05H  WZ,WC    ;ZC field set   111 
  qIF_Z      qXOR 0AH,#05H  WC,WZ    ;ZC field set   111 
  qIF_Z      qXOR 0AH,#05H  NR       ;R field clr    000
             qXOR 0AH,#05H           
             qXOR 0AH,#05H  WC   ; Fixed with  push iterate flag,suffix into G_CDS?        
  qIF_ALWAYS qXOR 0AH,#05H 
  qIF_ALWAYS qXOR 0AH,#05H WZ
  qIF_ALWAYS qXOR 0AH,#05H WZ,WC
  qIF_ALWAYS qXOR 0AH,#05H WZ,WC,NR
  
  ; field checks 
  	dd	(001b shl 23)  ; default
  	dd	(101b shl 23)  ; wz
  	dd	(111b shl 23)  ; wz wc
  	dd	(000b shl 23)  ; nr
  
  
• cgracey Posts: 14,280

  2016-09-21 21:57

  Looks really neat, jmg. I wish I had time to learn about it right now.
• jmg Posts: 15,191

  2016-09-21 22:40

  cgracey wrote: »

  Looks really neat, jmg. I wish I had time to learn about it right now.

  
  So far it looks promising, the idea is to find a good Macro Assembler, so you can focus on other, more critical things instead...
  
• Cluso99 Posts: 18,069

  2016-09-21 22:59

  jmg,
  I would like to see a short program using just a couple of P1 instructions, and the results of the compilation to get a better feel for it.
  
  If the preliminaries look good, then there are probably enough of us to write macros for each instruction once we have a complete macro for one instruction (ie including labels, condcodes, etc).
• Dave Hein Posts: 6,347

  2016-09-21 23:04

  The P2 pointer syntax may be a bit of a challenge for the assembler. It will also be interesting to see if it can automatically figure out which JMP code to use under different conditions. And then there's the differences in hub versus cog addressing, and when to use relative addresses versus absolute addresses.
• jmg Posts: 15,191

  2016-09-22 01:37

  Cluso99 wrote: »

  jmg,
  I would like to see a short program using just a couple of P1 instructions, and the results of the compilation to get a better feel for it.
  
  If the preliminaries look good, then there are probably enough of us to write macros for each instruction once we have a complete macro for one instruction (ie including labels, condcodes, etc).

  
  The code above covers Immediate, and the condition code IF_ prefix, and ADD and SUB and will produce a binary file.
  Not sure about any endian issues yet ?
  
  I'm looking at suffix handling for WC,WZ,NR, and have a limited version of that working.
  Need some tips from a fasmg expert to merge the prefix & suffix code ...
  
  
  
• jmg Posts: 15,191

  2016-09-22 20:58 edited 2016-09-22 21:05

  Cluso99 wrote: »

  jmg,
  I would like to see a short program using just a couple of P1 instructions, and the results of the compilation to get a better feel for it.

  With some help, I've now sorted suffix variants too, with white-space and updated post #2 with example test coverage.
  
  Speed checks :
  ; Listing ON -> 2 passes, 2.4 seconds, 16547 bytes, 266k LST
  ; Listing OFF -> 2 passes, 0.4 seconds, 16547 bytes, 80 byte LST
  
  
• jmg Posts: 15,191

  2016-10-18 05:50 edited 2016-10-18 19:46

  more:
  fasmg can export .hex via hex.inc macros, but that's slow, and seems to interact with other macros.
  
  addit: New versions of Listing.inc and hex.inc followed this slow/interact report, and they are now (much) faster & interaction is fixed.
  see: https://board.flatassembler.net/topic.php?p=191119#191119
  
  
  Alternative HEX and DB outputs :
  Some web searching finds srec_cat
  
  Download https://sourceforge.net/projects/srecord/files/srecord-win32/
  then run as

  srec_cat P8X32_PropMacros.bin -binary -o P8X32_PropMacros.hex -intel
  

  and srec_cat can also do this, which is an ASM syntax DB list variant.

  srec_cat P8X32_PropMacros.bin -binary -o P8X32_PropMacros.DB -ASM 
  
  (strangely, -ASM options seems to fail, but default decimal is tolerable.)
  
  creates
  ; http://srecord.sourceforge.net/
          DB      251,247,35,246,253,251,35,246,37,38,128,255,40,128,102,253
  ...
          DB      111,3,14,60,111,3,14,60,111,3,14,60,111
  ; upper bound = 0x4088
  ; lower bound = 0x0000
  ; length =      0x4088
  

  
  which can be INCLUDED into OnePin MCU ASM source, and it will auto-append the P2 code, to the small loader stub.
  
  
• jmg Posts: 15,191

  2017-02-21 19:38

  I'll bump this Assembler thread, with an assembler comment from the now-closed thread...
  

  David Betz wrote:

  ....Yup. That's one of the main reasons I haven't done anything on P2 for quite some time. .I need the instruction set and particularly the instruction encodings to remain stable. I also need the assembler syntax to be stable. I don't want to spend time modifying GAS for P2 only to find that a new assembler syntax gets invented and people start complaining that GAS doesn't match PASM.

  
  but I think we already have 'GAS does not match PASM' problems ?
  This is more an opportunity to bring PASM more into line with other assemblers ?
  
• David Betz Posts: 14,519

  2017-02-21 19:59

  jmg wrote: »

  I'll bump this Assembler thread, with an assembler comment from the now-closed thread...
  

  show previous quotes

  David Betz wrote:

  ....Yup. That's one of the main reasons I haven't done anything on P2 for quite some time. .I need the instruction set and particularly the instruction encodings to remain stable. I also need the assembler syntax to be stable. I don't want to spend time modifying GAS for P2 only to find that a new assembler syntax gets invented and people start complaining that GAS doesn't match PASM.

  
  but I think we already have 'GAS does not match PASM' problems ?
  This is more an opportunity to bring PASM more into line with other assemblers ?

  Yes, we already have those problems with the P1 version of GAS. I'd like to avoid that as much as possible with P2. It's a big pain to rewrite the instruction parsers inside of GAS. I already did it once for P2-hot and pretty much none of that code will transfer over so I don't want to have to write it more than once more.