Compiler (DBPro) - GameCreators Forum

Author

Message

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 10th Dec 2004 04:54 Edited at: 20th Dec 2004 18:44

Link

Simple interpreter / compiler type thing I did this evening.
Nothing too exciting yet - just outputs asm for mathematical expressions.

For example type

1+2-3*4/5

and see the screen fill with ASM

Tomorrow I'll try and do parenthesis (brackets).

rem Basic interpreter
#constant go goSub
global InputPos
global Input$
global Look$

` Start parsing program
go Program

end

`Keep getting lines of input
Program:

do

go Init
    go MainProg

loop

return

` --------------------------------------
` Get a code string and start prog
` --------------------------------------
Init:

`Get an input
  InputPos = 0
  input ":",input$

`Get lookahead char
  GetChar()

return

` --------------------------------------
` Parse a program
` --------------------------------------
MainProg:

`Beginning of program
  go Expression

return

` --------------------------------------
` Parse a term
` --------------------------------------
Term:

go Factor

getchar()

while Look$ = "*" or Look$ = "/"

Emit("MOVE D0,-(SP)")

done=0
    if Look$ = "*" then go Multiply : done=1
    if Look$ = "/" then go Divide : done=1
    if done=0 then Expected("* or divide operator")

` Get next char for next pass
    Getchar()

endwhile

return

` --------------------------------------
` Parse a factor
` --------------------------------------
Factor:

Emit("MOVE #" + str$(GetNum()) + ",D0")

return

` --------------------------------------
` Parse an expression
` --------------------------------------
Expression:

go Term

`GetChar()

while Look$ = "+" or Look$ = "-"

Emit("MOVE D0,-(SP)")

done=0
    if Look$ = "+" then go Addit : done=1
    if Look$ = "-" then go Subtract : done=1
    if done=0 then Expected("+/- operator")

` Get next char for next pass
    Getchar()

endwhile

return

Addit:

Match("+")

go Term

Emit("ADD (SP)+,D0")

return

Subtract:

Match("-")

go Term

Emit("SUB (SP)+,D0")
  Emit("NEG D0")

return

Multiply:

Match("*")
  go Factor
  Emit("MULS (SP)+,D0")

return

Divide:

Match("/")
  go Factor
  Emit("MOVE (SP)+,D1")
  Emit("DIVS D1,D0")

return

function Emit(string$)
  `  print "["+look$+"] - "+string$
  print string$
endfunction

function Debug(info$)
  `print info$
endfunction
` --------------------------------------
` Takes a char and matches it to the input
` if it doesn't match then put an error in
` --------------------------------------
function Match(char$)
  Debug("Match called with '"+char$+"'")
  if Look$ = char$
    Debug("matches")
    GetChar()
  else
    abort("'"+char$ + "' expected")
  endif
endfunction

` --------------------------------------
` Get character from input stream
` --------------------------------------
function GetChar()
  inc InputPos
  Look$ = mid$(Input$,InputPos)
  Debug("new look: "+look$)

endfunction

` --------------------------------------
` Gets a number from input stream
` --------------------------------------
function GetNum()
  num = val(Look$)
endfunction num

` --------------------------------------
` Abort Program
` --------------------------------------
function Abort(error$)
  Error(error$)

print "Any key to exit."
  wait key
end
endfunction

`report error
function Error(errstring$)
  ink rgb(255,0,0),0
  print "Err: ";errstring$
endfunction

function Expected(string$)
  Error("Expected "+string$)
endfunction

+ Code Snippet

rem Basic interpreter
#constant go goSub
global InputPos
global Input$
global Look$

` Start parsing program
go Program

end

`Keep getting lines of input
Program:

  do

    go Init
    go MainProg

  loop

return

` --------------------------------------
` Get a code string and start prog
` --------------------------------------
Init:

  `Get an input
  InputPos = 0
  input ":",input$

  `Get lookahead char
  GetChar()

return

` --------------------------------------
` Parse a program
` --------------------------------------
MainProg:

  `Beginning of program
  go Expression

return

` --------------------------------------
` Parse a term
` --------------------------------------
Term:

  go Factor

  getchar()

  while Look$ = "*" or Look$ = "/"

    Emit("MOVE D0,-(SP)")

    done=0
    if Look$ = "*" then go Multiply : done=1
    if Look$ = "/" then go Divide : done=1
    if done=0 then Expected("* or divide operator")

    ` Get next char for next pass
    Getchar()

  endwhile

return

` --------------------------------------
` Parse a factor
` --------------------------------------
Factor:

  Emit("MOVE #" + str$(GetNum()) + ",D0")

return

` --------------------------------------
` Parse an expression
` --------------------------------------
Expression:

  go Term

  `GetChar()

  while Look$ = "+" or Look$ = "-"

    Emit("MOVE D0,-(SP)")

    done=0
    if Look$ = "+" then go Addit : done=1
    if Look$ = "-" then go Subtract : done=1
    if done=0 then Expected("+/- operator")

    ` Get next char for next pass
    Getchar()

  endwhile

return

Addit:

  Match("+")

  go Term

  Emit("ADD (SP)+,D0")

return

Subtract:

  Match("-")

  go Term

  Emit("SUB (SP)+,D0")
  Emit("NEG D0")

return

Multiply:

  Match("*")
  go Factor
  Emit("MULS (SP)+,D0")

return

Divide:

  Match("/")
  go Factor
  Emit("MOVE (SP)+,D1")
  Emit("DIVS D1,D0")

return



function Emit(string$)
  `  print "["+look$+"] - "+string$
  print string$
endfunction

function Debug(info$)
  `print info$
endfunction
` --------------------------------------
` Takes a char and matches it to the input
` if it doesn't match then put an error in
` --------------------------------------
function Match(char$)
  Debug("Match called with '"+char$+"'")
  if Look$ = char$
    Debug("matches")
    GetChar()
  else
    abort("'"+char$ + "' expected")
  endif
endfunction

` --------------------------------------
` Get character from input stream
` --------------------------------------
function GetChar()
  inc InputPos
  Look$ = mid$(Input$,InputPos)
  Debug("new look: "+look$)

endfunction

` --------------------------------------
` Gets a number from input stream
` --------------------------------------
function GetNum()
  num = val(Look$)
endfunction num


` --------------------------------------
` Abort Program
` --------------------------------------
function Abort(error$)
  Error(error$)

  print "Any key to exit."
  wait key
end
endfunction

`report error
function Error(errstring$)
  ink rgb(255,0,0),0
  print "Err: ";errstring$
endfunction

function Expected(string$)
  Error("Expected "+string$)
endfunction

Have fun

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

Torrey

20

Years of Service

User Offline

Joined: 20th Aug 2004

Location: New Jersey

Posted: 10th Dec 2004 12:53 Edited at: 10th Dec 2004 12:55

Link

Don't forget in ASM, MOVE is really MOV.

Quick snippet:

; Mapping our executable into memory for manipulation
; and execution after converting
xor edx,edx
push edx
push edx
push 0
push 0xf001f
push eax
call [MapViewOfFile]

mov [mAddress],eax

call vpRoutine

Back to top

Profile PM Email

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 10th Dec 2004 15:45

Link

Quote: "Don't forget in ASM, MOVE is really MOV."

Not in my version

The 68k assembler I have uses move, move.l etc.

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

Shadow Robert

22

Years of Service

User Offline

Joined: 22nd Sep 2002

Location: Hertfordshire, England

Posted: 10th Dec 2004 17:43

Link

Heh, I believe we'd gone through this once

Still unsure why you'd deliberately start on an obsolete system.

Back to top

Profile PM Email Website

empty

22

Years of Service

User Offline

Joined: 26th Aug 2002

Location: 3 boats down from the candy

Posted: 10th Dec 2004 18:29 Edited at: 10th Dec 2004 18:30

Link

Nice adaption of the Pascal tut.

Quote: "Still unsure why you'd deliberately start on an obsolete system."

Maybe because the basic principles are the same on "modern" systems? As long as CPUs (and even most VMs) are stack based, the methods are the pretty much the same.

Play Nice! Play Basic!
Version 1.02 available now!

Back to top

Profile PM

Shadow Robert

22

Years of Service

User Offline

Joined: 22nd Sep 2002

Location: Hertfordshire, England

Posted: 10th Dec 2004 18:39

Link

perhaps, but the Pascal Assembler appears to have a few things that David isn't going to find in an Intel-Based Assembler.

i did suggest to him he visited the IRC room to find you, and talk to you about Pascal; as the only time i've developed for a 68k was through GAS, which appears not to have a few of the 'easy' commands that were giving him trouble.

Back to top

Profile PM Email Website

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 10th Dec 2004 21:53

Link

Quote: "Nice adaption of the Pascal tut. "

Yes

I wanted to recreate it from scratch, so I fully understood the concepts.

Raven - Whatever ASM I output doesn't really matter, because I can easily change what's outputted

Empty - I've been having trouble getting my asembler to accept this:

LEA variablename(PC),D0

It says unidentified identifier

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

empty

22

Years of Service

User Offline

Joined: 26th Aug 2002

Location: 3 boats down from the candy

Posted: 11th Dec 2004 00:46

Link

Quote: "Empty - I've been having trouble getting my asembler to accept this:"

Which assembler is it and can you post the rest of the code?

Play Nice! Play Basic!
Version 1.02 available now!

Back to top

Profile PM

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 11th Dec 2004 01:10

Link

I'm using one called easy 68k - it's a 68000 assembler I think - the same type the pascal tutorial uses.

Here's the full code, its a translated version of

var = 2 + 3

+ Code Snippet

move.l #2,D0
move.l D0,-(SP)
move.l #3,D0
add (SP)+,D0
lea var(PC),A0
move.l D0,(A0)

I assume the last two store D0 in the variable "var" - but it gives an error on the LEA line. I assume that's because I have to define it first?

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

empty

22

Years of Service

User Offline

Joined: 26th Aug 2002

Location: 3 boats down from the candy

Posted: 11th Dec 2004 04:09

Link

Yeah, lea var(PC), A0 stores the address of var (which needs to be defined, var: ds.l 1) in A0. Then Move.l D0, (A0) moves the contents of D0 to the address stored in A0.

Play Nice! Play Basic!
Version 1.02 available now!

Back to top

Profile PM

blanky

20

Years of Service

User Offline

Joined: 3rd Aug 2004

Location: ./

Posted: 20th Dec 2004 01:18

Link

The closest I've got to Assembler is failing to write a CHIP-8 emulator

(It works, but I get randomized dots on the screen which look like nothing human, which means I probably got one of the key functions wrong. Oh well.)

Back to top

Profile PM Email Website

Kevin Picone

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: Australia

Posted: 20th Dec 2004 16:14 Edited at: 10th Apr 2022 16:44

Link

David,

The parser has defined the label "Var", which the assmebly is expecting to be deifned and PC realtive. The label can appear before or after it's use, but.. Since it's PC relative it must be within a 16bit range (signed) of where it's used.

Since the assembler will not store the absolute address of this variable in memory, but it's offset from where the PC is currectly executing (the LEA opcode), it creates the A0 pointer by basically going A0 = PC + offsetToVariable.

if it were an absolute LEA you could just

Lea.l MyVariable,a0

So you could (define the variables first)
=======================

Bra Start ; skip the data segment

MyVariable: Dc.l 0 ; Define space for a long

Start:
move.l #2,D0
move.l D0,-(SP)
move.l #3,D0
add (SP)+,D0
lea MyVariable(PC),A0 ; << load relative address to MyVariable into A0
move.l D0,(A0)

brk ; force an exception (halt program)

or (define it after),
==========================

move.l #2,D0
move.l D0,-(SP)
move.l #3,D0
add (SP)+,D0
lea MyVariable(PC),A0 ; << load relative address to MyVariable into A0
move.l D0,(A0)

brk ; force an exception (halt program)

MyVariable: Dc.l 0 ; Define space for a long

A better solution would be use a format as RS structures for variables.. This is most compiler like SASC on the amiga do. This means that a register is used as the base addfress of the variable stack

I.e.

Variables:
Blk.l NumberOfVariables ; Define Space for the cleared block LONGS, at the label Variable

At the start of your code, you LEA this into an address register, like A5, A6 say

Lea.l Variables,A5

Next whenever you generate variable access, rather than LEA'ing the absolute address of the variable and then doing a move (LEA's are around 6->8 clock cycles, and two register moves to memory are a bit less, probably 4 clocks.. ).. you can use a fixed 16bit offset to the A5 register.. The compiler just needs to translate that variable name, into it's index on the variable stack.

I.e.
MOve.l #10,(a5) ; move 10 into the first variabe

Move.l #20,1*4(a5( ; move th 20 into the second variable

Etc..

rather than

Lea.l MyVariable(pc),a0
move.l #10,(a0)
Lea.l MyOtherVariable(pc),a0
move.l #20,(a0)

On the 68000 and 68010 there are some opcodes that are not supported in this format, but ya get that

68020 and above support up to 32bit offsets..

LearnToCode with Kevin Picone at www.PlayBASIC.com

PlayBASIC To HTML5/WEB - Convert PlayBASIC To Machine Code

Back to top

Profile PM Website

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 20th Dec 2004 18:44

Link

Quote: "The parser has defined the label "Var", which the assmebly is expecting to be deifned and PC realtive. The label can appear before or after it's use, but.. Since it's PC relative it must be within a 16bit range (signed) of where it's used. "

Yes I realised that afterwards! When / If I ever finish this, any declarations would be made at the beginning of the program and translted into ASM declarations.

Thanks for the advice anyway, I'll have a look at it.

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

Kevin Picone

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: Australia

Posted: 20th Dec 2004 21:51

Link

Well you could do it that way, but then each variable needs to be explicitly declared ahead of time in the source, or the pre-processor has to find declarations. Which can get really messy.

Also, really need to watch the operation width..

I.e.

add (SP)+,D0

should be explicit,

add.l (sp)+,d0

For stuff like brackets, the parser is just being made to focus it's attention upon the characters within the bracket pair, until the there's a firm result (could be syntax error also.. ect) . The result is then exported from the brackets and re-embedded in the expression and the process continues.. until all that remains is the result

While working on strings might be easier to get the head around, i'd really suggest moving to tokens.. It's allows for pre & post processing of the expressions..

Kevin Picone
[url]www.underwaredesign.com[/url]
Play Nice! Play Basic - Next Generation Basic (Release V1.05 Out Now)

Back to top

Profile PM Website

David T

Retired Moderator

22

Years of Service

User Offline

Joined: 27th Aug 2002

Location: England

Posted: 20th Dec 2004 23:04

Link

Quote: "but then each variable needs to be explicitly declared ahead of time"

That's what I'm heading for

Get 15 new commands, all the date / time commands left out of DBPro for free!
DOWNLOAD PLUGINS HERE: http://www.davidtattersall.me.uk/ and select "DarkBasic"

Back to top

Profile PM Email Website

Sorry your browser is not supported!

Code Snippets / Compiler (DBPro)