[Coco] Assembly .BIN -> .CCC/.ROM = SOLVED!

[Chad H]

I will check this after work when I get home..might save couple of bytes.  Not sure I understand the JMP [2,X] at the end.  X being the offset in the ROM image, I don't want to execute it. I want to execute the address that is stored there (.bin postamble). I will try and see.

On Sep 4, 2014 9:31 AM, Johann Klasek <johann+coco at klasek.at> wrote:
On Wed, Sep 03, 2014 at 09:47:36PM -0500, Chad H wrote:
> Ahh... when I was looking in the Rainbow IDE HELP > 6809 refernce at the
> opcode table and saw the DECD, I failed to see the note that '*' represented
> a 6309 code...oops.  Anywhoo, I FINALLY got this bugger going perfectly!! :)
> Thanks to William Astle, Johann Klasek and Tormod Volden, you guys all made
> points that helped me debug this ..umm..bugger.
>

Just for elegance (efficiency also), if someone bother  ... ;)
I would change Y <-> U as William pointed out already:

>       org     49152           CoCo and compatibles map in ROM Paks here
>
>       * LOADER
> BININ LDX     #BINLOD         INIT XFER DATA ADDRESS OFFSET
> CHKBLK        LDA     ,X+             GET BLOCK TYPE BYTE (00 = PREAMBLE,
> 255=POSTAMBLE)
>       BNE     ENDBIN          IF <>0 THEN MUST BE END OF .BIN DATA
> (POSTAMBLE)
>       LDU     ,X++            GET BLOCK LENGTH(U)

-       LDU     ,X++            GET BLOCK LENGTH(U)
+       LDY     ,X++            GET BLOCK LENGTH(U)

>       LDY     ,X++            GET BLOCK START ADDRESS(Y)

-       LDY     ,X++            GET BLOCK START ADDRESS(Y)
+       LDU     ,X++            GET BLOCK START ADDRESS(Y)

> XFER  LDA     ,X+             GET SOURCE BYTE(A) FROM X
>       STA     ,Y+             PUT BYTE(A) AT Y

-       STA     ,Y+             PUT BYTE(A) AT Y
+       STA     ,U+             PUT BYTE(A) AT Y

>       LEAU    -1,U            MOVED BLOCK?
>       CMPU    #0

-       LEAU    -1,U            MOVED BLOCK?
-       CMPU    #0
+       LEAY    -1,Y            MOVED BLOCK?

>       BNE     XFER            NO
>       JMP     CHKBLK          CHECK NEXT BLOCK
> ENDBIN        LDU     ,X++            GET BLOCK LENGTH (0000)
>       LDU     ,X++            GET EXECUTION ADDRESS(U)

Without the skipping read (if X is not expected to point after
the execution address) ...

- ENDBIN        LDU     ,X++            GET BLOCK LENGTH (0000)
-       LDU     ,X++            GET EXECUTION ADDRESS(U)
+ ENDBIN        LDU     2,X     GET EXECUTION ADDRESS(U)


>       JMP     [,U]

Is this really meant in this way? Register U contains already
the address where the execution should proceed, isn't it?
This would fetch the 2 bytes U is pointing at, which is the
effective address here (which is already the code itself).
If U already contains the address, simply jump there with

        JMP     ,U

Same as the original implemention does:

        LDY     ,X++            GET EXECUTION ADRESS(Y)
        STY     EXECJP
        JMP     [EXECJP]
or
        LDY     ,X++
        JMP     ,Y
or
        JMP     [,X]

Back to the version above:
This could be shorten further (without reading the block length)
with

ENDBIN  JMP     [2,X]

(replacing both LDUs and the JMP)

In addition I prefer BRA instead of JMP (to keep it relocatable, just
in case).

Putting altogether:

        org     49152           CoCo and compatibles map in ROM Paks here

        * LOADER
BININ   LDX     #BINLOD         INIT XFER DATA ADDRESS OFFSET
CHKBLK  LDA     ,X+             GET BLOCK TYPE BYTE (00 = PREAMBLE, 255=POSTAMBLE)
        BNE     ENDBIN          IF <>0 THEN MUST BE END OF .BIN DATA (POSTAMBLE)
        LDY     ,X++            GET BLOCK LENGTH(Y)
        LDU     ,X++            GET BLOCK START ADDRESS(U)
XFER    LDA     ,X+             GET SOURCE BYTE(A) FROM X
        STA     ,U+             PUT BYTE(A) AT U
        LEAY    -1,Y            MOVED BLOCK?
        BNE     XFER            NO
        BRA     CHKBLK          CHECK NEXT BLOCK
ENDBIN  JMP     [2,X]           SKIP BLOCK LENGTH (0000)
                                AND JUMP TO EXECUTION ADDRESS
BINLOD  FCB     255


Anything wrong with this?


Johann


--
Coco mailing list
Coco at maltedmedia.com
http://five.pairlist.net/mailman/listinfo/coco