Sonic & Knuckles Collection C port

Is there any more information on which memory regions are used for what by the Special Stage? Here's what I have so far:

• A400-A4FF - collision response list
• A500-???? - temporary data while checking if we tagged a group of spheres?
• F100-F500 - layout of the current stage

I'm scratching my head as what's happening in the equivalent of the disassembly's sub_9EBC label.

 

I might have bitten off more than I can chew by trying to port the Special Stage. I'm a good decompiler/porter, but as I'm not that good at reverse engineering I depend on existing documentation. There's too little for the Special Stage.

After some refactoring, I took a look at the code for Sonic 3's title screen. Sonic & Knuckles Collection has its code entirely separate from Sonic 3 & Knuckles's, but as expected they're very similar. They even went as far as to duplicate mapping and animation data, which I have verified are identical.

Not wanting to repeat this unnecessary duplication, I decided to merge Sonic 3's title screen code with Sonic 3 & Knuckles.

Some observations:

• Unlike the other title screens, Sonic 3's doesn't bother hiding the trademark (TM) symbol if the game is running on a domestic system.
• Sonic 3 & Knuckles polls both controllers for the Start button. Sonic 3 only polls the first controller.

 

Sonic 3's level select has also been ported! This time I didn't merge all the code, but only that of the subroutines. Merging the main routines would have been a bit too messy.

I had never taken a close look at this level select before due to never having succeeded in entering the cheat code for it, but the zone icons are those of Sonic 2!

Another curiosity is that the cheat code to get all emeralds has two possible results depending on the value of register d2. If d2 is zero, you get 15 continues. If it's non-zero, you get all emeralds. But d2 is always 1.

 

I'm still stumped by Sonic & Knuckles's title screen.

@Devon could there still be something wrong with my mapdevr port? Or maybe my port of moduled unlze?

 

Looking much better!



Something must be off with the code handling the animation, because Sonic's left hand doesn't display properly, and Knuckles's hands don't animate at all. Still, progress!

 

I've finished most of Sonic & Knuckles's title screen!



I say "most" because the phaseout effect on the logo doesn't work, and part of the bottom of the background shows at the top when Sonic lands.

The character selection roulette works, but I'm not sure about the implementation. To decide whether the Knuckles name tag has to be at the front, the variable that keeps track of the rotation is decremented with 160, and it checks if the result is negative. This seems to rely on the underflow behaviour of a signed char. So I tried to check if the variable is higher than 32 instead, thinking that decrementing 0 by 128 would cause an underflow, making the number positive again.

However, during testing I saw that when the rotation has stopped at Knuckles's name tag, the rotation variable is -128. So I added an extra check for that.

After this, I wasn't satisfied, because it didn't seem that the two name tags were equally highlighted while at the front during the rotation, so I changed the value of 32 by 64.

Any ideas or corrections for this logic are welcome.

 

Many pages back, I was implementing a common function that did DPLC. I'm looking at the Special Stage code again, and this part at the end of the player code looks similar to it:

Code (ASM):

1. SStage_PLCLoad_91A2:
2.        moveq    #0,d0
3.        move.b    mapping_frame(a0),d0
4.        cmp.b    $3A(a0),d0
5.        beq.s    locret_91E6
6.        move.b    d0,$3A(a0)
7.        add.w    d0,d0
8.        adda.w    (a2,d0.w),a2
9.        move.w    (a2)+,d5
10.        subq.w    #1,d5
11.        bmi.s    locret_91E6
12.  
13. loc_91BE:
14.        moveq    #0,d1
15.        move.w    (a2)+,d1
16.        move.w    d1,d3
17.        lsr.w    #8,d3
18.        andi.w    #$F0,d3
19.        addi.w    #$10,d3
20.        andi.w    #$FFF,d1
21.        lsl.l    #5,d1
22.        add.l    d6,d1
23.        move.w    d4,d2
24.        add.w    d3,d4
25.        add.w    d3,d4
26.        jsr    (Add_To_DMA_Queue).l
27.        dbf    d5,loc_91BE

However, it seems to do things a bit differently. The first two bytes still contain the count, and I think the instruction follows. But in contrast to the function I ported, the instruction also needs to be byte-swapped in the PC version. After that, I don't know what it's doing. The left shifting by 5 bits makes me think that's where a vram address is extracted.

 

Thanks! I've translated it to the following C code:

Code (C):

1.  
2.   if (p_actwk->patno != SPRITE_STATUS_UCHAR(p_actwk, 58)) {
3.     SPRITE_STATUS_UCHAR(p_actwk, 58) = p_actwk->patno;
4.     unsigned short* p_wpat = p_wrtpat[p_actwk->patno];
5.     unsigned short cnt = *p_wpat++;
6.    
7.     for (int i = 0; i < cnt; ++i) {
8.       unsigned short instruction = *p_wpat++;
9.       unsigned short words = ((instruction >> 8) & 0xF0) + 16;
10.       unsigned char* p_source = p_sourcebase + (instruction & 0xFFF) * 32;
11.       copy_data_vram(p_source, vram_adr, words);
12.       vram_adr += words * 2;
13.     }
14.   }

By the way, the reason I got back to Special Stage code is because I found some documentation on its inner workings on Sonic 3 Unlocked.

Still not confident I can get this working, but we'll see how far I get.