Best way to write stuff on screen

[E-Penguin]

I couldn't get vasm to output the same output as was required by Blitz so I got out the hex editor and the motorola 68000 datasheet and edited the op code.

It now successfully adds a long's worth of BCD digits.

Let's just say this has been a learning experience.

For completeness I'm going to do the same for sbcd and nbcd then publish the library here.

[clenched]

VASM did OK using this procedure. I'm stunned the Blitz assembler is so fouled up.

** vasmm68k_mot -devpac -Fhunkexe -kick1hunks -nosym -o \bl.obj \bl.asm
** bl stands for blitz library
** Blitz wants executable despite the .obj name
** Put in otherlibs then run makedeflibs

 INCLUDE "\blitz2\Blitz.i"
 libheader 50,0,0,0,0
 astatement
 args long,long,byte
 libs
 subs _AddBCD,0,0
 name "AddBCD","Num1,Num2,Size"
 libfin

_AddBCD:
  MOVE.l d0,a0 ; num1 in d0 to a0
  ADDA d2,a0  ;  increment a0 by the length of the string
  MOVE.l d1,a1 ; num2 in d1 to a1
  ADDA d2,a1  ; increment a1 by the length of the string
  SUBI #1,d2 ; decrement the string length

  MOVE #$04,CCR ; clear the flags
_AddBCD_loop:
  ABCD -(A0),-(A1) ; predecrement addresses, A0 = A0 + A1
  DBRA D2,_AddBCD_loop ; decrement loop counter and go again if not zero
  RTS

[E-Penguin]

Nice. Thanks for the tip; certainly simpler than going at it with a hex editor! Where did you find Blitz.i?

[clenched]

It's easier to paste here than remember where it came from.
EDIT 1&3: Archive URL where blitz.i originates in ZONE! http://aminet.net/dev/basic/BlitzASMInclud.lha
EDIT2: These BCD bugs also exist in Amiblitz2. Double checked with WinUAE debugger and Blitz's own.

[clenched]

Blitz macros for inline BCD instructions:
Arg 1 - Mode Address or Data register. Any number would do, $a(ddress) $d(ata) just makes it easier to tell which is used. ie number is odd or even.
Arg 2,3 - Register number 0-7.

Macro _ABCD
 Dc.w `1 LSL 3-1 AND 8 OR `2 OR `3 LSL 9 OR $c100
End Macro

Macro _SBCD
 Dc.w `1 LSL 3-1 AND 8 OR `2 OR `3 LSL 9 OR $8100
End Macro

TST $100
JMP over
ABCD -(a0),-(a1)
ABCD d2,d5
SBCD -(a6),-(a3)
SBCD d7,d1
RESET
!_ABCD{$a,0,1}
!_ABCD{$d,2,5}
!_SBCD{$a,6,3}
!_SBCD{$d,7,1}
over:
Stop

00338F6A 4ef9 0033 8f82           JMP $00338f82
00338F70 c10d                     ABCD.B -(A5),-(A0)
00338F72 c505                     ABCD.B D5,D2
00338F74 8d0f                     SBCD.B -(A7),-(A6)
00338F76 8f01                     SBCD.B D1,D7
00338F78 4e70                     RESET 
00338F7A c308                     ABCD.B -(A0),-(A1)
00338F7C cb02                     ABCD.B D2,D5
00338F7E 870e                     SBCD.B -(A6),-(A3)
00338F80 8307                     SBCD.B D7,D1

[E-Penguin]

It's still getting the registers wrong, no?

00338F70 c10d                     ABCD.B -(A5),-(A0)

Whereas your first macro is a0, a1

*edit*

Aah your macros are fixing the assembler output by directly calculating the op code. That's very clever!

[clenched]

Same thing whittled down to two arguments. Looks a little less alien. The d's look almost real.

Macro _ABCD
 Dc.w {`1 BitChg 4 AND 16} LSR 1 OR {`1 AND 7} OR {`2 AND 7} LSL 9 OR $c100
End Macro

Macro _SBCD
 Dc.w {`1 BitChg 4 AND 16} LSR 1 OR {`1 AND 7} OR {`2 AND 7} LSL 9 OR $8100
End Macro

!_ABCD {$a0,$a1}
!_ABCD {$d2,$d5}
!_SBCD {$a6,$a3}
!_SBCD {$d7,$d1}

[clenched]

Edit Blitz2 as shown. Copy of Blitz2 is the original. This exchanges some bits and sets some to match similar ADDX,SUBX construction. Macro synthesizes source code to test the patch. If source is to be shared maybe better to stick to macros.

Comparing files C:\BLITZ2\Copy of Blitz2 and C:\BLITZ2\BLITZ2
00000F28: 90 0F
00000F29: 00 9F
000018F0: 90 0F
000018F1: 00 9F

EDIT: Streamlined to eliminate edit test.txt step.

; assemble: vasmm68k_mot -Fbin -o \test.txt \test.asm
; run test.txt in blitz
; assemble: vasmm68k_mot -Fbin \test.txt
; compare blitz.out and a.out, should be the same

        dc.b " rem",$0a
        dc.b 'l.l=WriteFile(0,"ram:blitz.out")',$0a
        dc.b "WriteMem 0,?strt,512",$0a
        dc.b "CloseFile 0",$0a
        dc.b "End",$0a
        dc.b " erem",$0a
        dc.b "strt:",$0a

test    macro
        dc.b "  abcd.b  -(a\<src>),-(a\<dst>)",$0a
        dc.b "  sbcd.b  -(a\<src>),-(a\<dst>)",$0a
        dc.b "  abcd.b  d\<src>,d\<dst>",$0a
        dc.b "  sbcd.b  d\<src>,d\<dst>",$0a
        endm

src     set     0
        rept    8
dst     set     0       
        rept    8
        test
dst     set     dst+1
        endr
src     set     src+1 
        endr

[E-Penguin]

It might be worth uploading that to aminet or similar as a patch, then hopefully it'll be widely adopted

[clenched]

Originally Posted by E-Penguin

It might be worth uploading that to aminet or similar as a patch, then hopefully it'll be widely adopted

It would be better if a Blitz developer took a look first. This opcode test along with trying to fool it e.g. abcd.w is the extent of testing. Meanwhile other software is coming apart at the seams. Assemble and immediately disassemble these w/ Action Replay III:
abcd.b -(a0),-(a1)
addx.b -(a0),-(a1)
Now check that against what WinUAE debugger says.