Effectus - New Atari cross-compiler (Alpha stage)

[Gury]

Thanks for the info! It comes from official Action! language documentation. I checked it out and it makes sense to me now. Ok, parameter issue is resolved, not it's up to me to make proper updates in Effectus.

[fujidude]

Keep in mind, Action by default will try to create storage for variables local to a procedure/function at the start of where it compiles that proc/func to. In such case, the the info passed to it get stored there and not to the CPU registers and page 0. I think it's the @ symbol that instructs Action not to set up storage and instead use the registers and page 0. I would have to look that up though.

[Gury]

I haven't read this thread for a while so I need to read all 9 pages ... what's the reason this is dropped in the new release "Custom ASM functions: register y removed from the parameter list" ?

 

Hi, here is new version of Effectus (0.2.2). I have already fixed the problem with machine language routine parameters last year, but I didn't publish new version yet. Now it handles parameters correctly, at least for examples I tested. The thing which does not work correctly yet is calling extra routines inside machine language routines, which will be checked too, of course.

 

Here is an example of new feature:

; CODE BLOCK WITH PARAMETERS EXAMPLE

PROC TEST=*(BYTE CURSOR,BACK,BORDER,X,Y,UPDOWN)
[$8E 710  ; BACKGROUND COLOR
 $8C 712  ; BORDER COLOR
 $8D 752  ; CURSOR VISIBILITY 
 $A5 $A5 $8D 755  ; CHARACTERS UPSIDE DOWN? 
 $A5 $A3 $8D 85 0  ; COLUMN FOR TEXT
 $A5 $A4 $8D 84 0  ; ROW FOR TEXT
 $60]

BYTE FUNC ADD8=*(BYTE ADD3,ADD4)
[
$D8      ; CLD
$18      ; CLC
$86 $B1  ; STX $B1
$65 $B1  ; ADC $B1
$85 $A0  ; STA $A0
$60]     ; RTS

CARD FUNC ADD16=*(CARD ADD1,ADD2)
[
$D8      ; CLD
$18      ; CLC
$86 $B1  ; STX $B1
$84 $B2  ; STY $B2
$65 $B2  ; ADC $B2
$85 $A0  ; STA $A0
$A5 $B1  ; LDA $B1
$65 $A3  ; ADC $A3
$85 $A1  ; STA $A1
$60]     ; RTS

PROC MAIN()

BYTE X1
CARD NUM

PUT(125)
TEST(1,20,30,8,8,4)
PRINTE("LOOK, I AM UPSIDE DOWN!")

PutE()
PutE()
X1=ADD8(100,50)
Print("100 + 50 = ")
PRINTBE(X1)

NUM=ADD16(1100,34900)
Print("1100 + 34900 = ")
PRINTCE(NUM)

RETURN

Greblus... Linux version please. Thank you!

 

Greetings

func_asm.xex

Edited January 24, 2016 by Gury

[greblus]

Here it is (with mads included)

effectus-0.2.2.tar.gz

[Gury]

Thank you!

[Gury]

Small update to version 0.2.3 here.

 

Download

 

Current version of Effectus is compiled with Free Pascal 2.6.4. You should use Mad Assembler 2.0.0 for proper cross-linking of Atari 8-bit executables.

 

Greetings

[greblus]

Hi Gury.

 

Here's Linux version.

 

W.

effectus-0.2.3.tar.gz

[Gury]

I have overslept

 

Thank you, greblus! It is uploaded.

[+Spaced Cowboy]

So I'm not sure if this forum is still being updated, but as I go through the code for effectus I think I've found a couple of bugs.

 

1) In core.pas, starting at line 475 (Arrays and pointer detection)

    for j := 1 to GVarCnt2 do begin
      if System.Pos(GVar2[j].VarType, 'T5T6T7T8T10') > 0 then Continue;
      n := System.Pos(UpperCase(GVar2[j].VarName), UpperCase(TextBuf[CurLine]));
      if (n > 0) and VarDeclCheck(TextBuf[CurLine]) then begin
        Buffer := Trim(TextBuf[CurLine]);
        
        // Comments
        if System.Pos(';', Buffer) > 0 then begin
          CodeBuf.Add('; ' + ExtractEx(Buffer, ';', 2, []));
          Buffer := ExtractEx(Buffer, ';', 1, []);
          n := System.Pos(UpperCase(VarTypes[j]) + ' ', UpperCase(Buffer));
          if (n < 1) then Continue; 
        end;

You can see that 'j' is iterating over the 'GVarCnt2' array, but then is used to reference VarTypes[j] on the third-to-last line. This looks like a cut/paste issue because the section immediately above (Scalar type variables) does iterate 'j' over VarTypes[] and has the same block of code. Perhaps the test for n ought instead to look like:

n := System.Pos(UpperCase(GVar2[j].VarType) + ' ', UpperCase(Buffer));

2) In common.pas, starting at line 191, we have:

          if (System.Pos('PROC ', UpperCase(TextBuf[CR_LF])) > 0) then begin
            Buffer := ExtractEx(TextBuf[CR_LF], 'PROC ', 2, []);
            if System.Pos('=', TextBuf[CR_LF]) > 0 then
              Buffer := ExtractEx(Buffer, '=', 1, [])
            else if System.Pos('=*', TextBuf[CR_LF]) > 0 then
              Buffer := ExtractEx(Buffer, '=*', 1, [])
            else begin
              Buffer := ExtractEx(Buffer, '(', 1, []);
            end;
            ProcBuf.Add('PROC' + Buffer);
          end {else begin}

It seems to me that the search for '=' on the third line will mean that the search for '=*' (on the fifth line) will never be called, because any occurence of '=*' would have been detected by the search for just '='. Perhaps the order these are tested for ought to be inverted ?

 

I'll also CC: to Gury's email address (I couldn't find anything on the effectus home page, but there's one on http://gury.atari8.info/my_projects.php)

Edited January 17, 2018 by Spaced Cowboy

[Gury]

Thanks, Spaced Cowboy, for pointing out an issue. I will look at it and make some other small updates, but anyway I plan to refactor and recode the whole project from scratch at some point. That's probably a small spoiler now

 

But I encourage you and everyone else to make tests and report bugs if you find them.

 

Greetings

[Gury]

Here is experimental version of Effectus with integrated MADS - Mad-Assembler v2.0.8 code, so no extra MADS executable is needed for program to run. I included MADS source code as unit for Effectus and implemented a procedure to call it. I will include sources in next release bundle version.

 

This version is also a minor bug-fix release with my and Spaced Cowboy's fixes. It may be that some example programs in original releases will not work, but all will be fixed later.

 

I hope Tebe will not mind I did this

effmads.zip

Edited February 16, 2019 by Gury

[Alfred]

Keep in mind, Action by default will try to create storage for variables local to a procedure/function at the start of where it compiles that proc/func to. In such case, the the info passed to it get stored there and not to the CPU registers and page 0. I think it's the @ symbol that instructs Action not to set up storage and instead use the registers and page 0. I would have to look that up though.

 

Not quite right. A/X/Y are always used to hold the first three bytes of parameter data. The rest are stored by the caller at $A3-$AF. Upon entry to the PROC/FUNC the internal procedure SARGS is called to copy the data from where the caller stored it to the local variables of the PROC/FUNC. Using the =* form for a procedure suppresses the call to SARGS and eliminates the initial JMP to the following instruction.

 

Code from the cartridge:

 

; handle procedure setup here

pla

bmi seg120 ; system proc

 

;

; IF IT'S NOT A 'SYSTEM' PROC THEN WE NEED TO GENERATE

; THE CODE TO COPY THE PASSED ARGS FROM ARGS TO THE

; BYTES JUST DECLARED.

;

; get beginning of arguments and

; save actual procedure address

lda #1 ; GO GET START ADDRESS

jsr cprop ; OF THE PROC

sta arg0

stx arg1

jsr gcdoff ; GET CURRENT QCODE

jsr stprop ; AND STORE IN SYMTAB

 

; get space for proc variable

lda #$4c ; JMP

jsr push1

jsr gcdoff ; fill in address

adc #2 ; SKIP THE 2 JMP BYTES

bcc seg090

inx

seg090 jsr push2 ; AND STORE IN QCODE

 

; qcode to transfer arguments to

; local frame

seg100 lda arbyts ; HOW MANY

beq seg140 ; no arguments

cmp #3

bcs seg130

cmp #2

lda #$8d ; STA addr16

ldx arg0

ldy arg1

bcc seg110

lda #$8e ; STX addr16

inx

bne seg110

iny

seg110 jsr push3 ; STORE THE CODE

dec arbyts

jmp seg100

 

seg120 jmp seg170 ; SKIP TO SYSTEM PROC

 

seg130 ldx #10 ; GENERATE THE JSR TO 'SARGS'

jsr jsrtbl

lda arg0 ; STORE THE BYTES

ldx arg1 ; INLINE AFTER THE

ldy arbyts ; 'JSR SARGS' CALL

dey

jsr push3

Edited February 16, 2019 by Alfred