Damn assembler

[socram8888]

Long time without asking anything.

Well, some people at a Spanish forum tought about creating a videogame using BasiEgaXorz, and I'm working as a "coprogrammer", doing some coding

Well, in order to make the game smaller, I'm trying to build a compression/decompression function in Assembler, using a RLE-4 (Run Length Encode with 4 bits of data). A real uncompressed tile looks like this:
11 11 11 12 11 11 11 14 11 11 11 14 11 11 11 14 11 11 11 16 11 11 11 12 11 11 11 12 11 11 11 12
And RLE-4 compressed is:
71 12 71 14 71 14 71 14 71 16 71 12 71 12 71 42

That is:
7 times 1
2 once
7 times 1
4 once
and so on

Here's the decompressing routine:
lea.l DummyTestData, a0
move.l #$FFFF00, a1

move.b #4, d4
clr.b d3
GetNext:
move.b (a0)+, d0
move.b d0, d1
lsr.b #4, d1
cmp.b #0, d1
bne.b NotZero
rts
NotZero:
move.b d0, d2
and.b #15, d2
Loop:
move.b d2, d5
lsl.b d4, d5
or.b d5, d3
cmp.b #0, d4
bne.b DoNotSave
move.b d3, (a1)+
clr.b d3
move.b #8, d4
DoNotSave:
subq.b #4, d4
subq.b #1, d1
cmp.b #0, d1
bne.b Loop
beq.b GetNext

DummyTestData:
dc.b $55, $11, $89, $98, $01
On EASy68K it works perfectly, but in Gens KMod it does strange things like adding 2 instead of 1 in "(a1)+", or setting a1 to the start value, etc...

I tried disabling the interruptions and so on, but i'm unable to get it working

PS: Of course I'm using "@label" on BEX instead of "label" and I'm loading the DummyTestData from a data file

[oompa loompa]

That sounds weird. Try another address instead of $FFFF000, like $FFF000 (or declare a variable and extract the address using &varptr) because the stack may overwrite.

And uhm, I also hope you have an RTS after "beq.b GetNext"

[socram8888]

Well, I have a BNE and a BEQ, so it must branch anyway (because if it's not different - BNE, then it must have the same value - BEQ)

I'll try with another location...

[oompa loompa]

I dunno if this will help either, but maybe try doing "lea.l DummyTestData(PC), a0", as in fetch the location relative to PC. I would also look at the hex output of the assembled program to see if the asmx didn't do anything funny

[Mairtrus]

You are using it as a User Defined subroutine? Because remember that BEX doesn't pass an offset as a parameter, unless you use VarPtr&, LblPtr&, etc... I don't remember right, but I think I too had problems trying to pass an offset directly to a An register (I had to move it to a Dn register, the to an An register)
Aditionally, are you using the RAM viewer to check the status of the decompressed data? Because maybe this is what is confusing you: sometimes, when accesing a large amount of contiguous bytes, seems that KMod swap the odd and even bytes in a same word, but only in the RAM viewer, not in the real RAM (you can check this with a peek+print loop).

[socram8888]

I'm using A6 as input data offset, because this is set to the data block after doing a "reload". And yes, that part works and it's really set to the dummy data

[socram8888]

Well, I tried using another location (FFF000) and it's still glitchy :/

/offtopic: Does anyone how does the Gens KMod debugging works? I want to send to Messages some debugging info, but writting data to VDP register 30 does not work :S
move.w #$9E35, d0
move.l #$C00004, a6
move.w d0, (a6)
unl:
bra unl

[oompa loompa]

Mar 3, 2011 8:27:49 GMT -5 socram8888 said:

I'm using A6 as input data offset, because this is set to the data block after doing a "reload". And yes, that part works and it's really set to the dummy data

Oh, try using another address register. I don't remember what A6 does - I think this stores a pointer to the heap, but BasiEgaXorz loves to change this variable on the fly . I think address registers A3-A7 are taken by the compiler, so you only have A0-A2 to play with

[socram8888]

A6 is used as data pointer. Anyway, I'm copying it before doing anything to A0

I'll try tomorrow using a loop of peek/prints :/

[socram8888]

Finally got it working. It was that byte-swapping error, but using Peek/prints shows everything ok