Mortal Kombat II AGA Tech Test

[jotd]

I never used attached sprites, not on ECS and even less on AGA. What I know is that non-attached sprites have a separate palette, that would be nice to experiment with AGA attached sprites, see if 2 characters of 128 pixels each could be displayed using all sprites (like the demo pretends but we don't see 2 characters at once).

If this is possible, then a game like MK would be completely possible, even a transcode from the original. I'd love to get my hands on the original MK source code...

[alexh]

Originally Posted by jotd

I'd love to get my hands on the original MK source code...

The Amiga MK1 source code? I've only ever seen the Arcade MK2 & N64/Windows/PS1/Arcade "MK Trilogy" source floating around.

One of the programmers, Richard Costello was in for Retro Gamer, Issue 243, 16 February 2023, p40 so he's still around.

[roondar]

Originally Posted by Old_Bob

Would that be 15 colours, divided between both fighters if they're being used in attached mode? Or, 15 each? I know we can switch colour banks for the sprites on AGA, but without checking I can't recall the specifics. If they're shared, it's not much of an visual improvement over the version we have, already.

In either case, it would reduce the processing overhead for the graphics engine by quite a bit, of course, as well as the memory requirement. For an A1200 or CD32 with Chip RAM only, it would probably be the way to go.

B

AGA attached Sprites all share the same palette, though you can select where in the in colour palette the Sprite colours are stored* rather than always being stuck with colours 17-31. So, they'd be limited to 15 colours shared between both fighters.

Unattached Sprites can still only use 3 colours each, but because odd and even Sprite colours can have their seperate banks, all 8 Sprites can have their own 3 colour palette (the attached Sprites all use the 'odd Sprites' palette bank).

*) With some limits, but essentially you can choose each '16 colour aligned block' for use with them.

All in all, you can certainly use HW Sprites for characters in MK2 AGA, but it'll look a lot less colourful than the demo you've provided here.

[jotd]

There's another possibility: superimpose 4 HW sprites to get 12 independent colors per character (more with copper dynamic color change), but the size will be 64 max unless you reserve 3 colors for the extreme left/right of the sprites, in which case you could go wider... or complement wide moves with blitter.

I've done such a thing on Xevious port: the big ship is 96 pixels wide. I've used 4 sprites which could display 6 different colors. The original ship uses 7 colors, I removed a shade somewhere and it's almost the same. And it allowed to use the 4 other sprites with completely independent colors (player ship also used 2 sprites, plus other specially colored enemies).

MK source code (amiga or original) would be great news. I can re-do things without the source but it takes a lot more time.

[ransom1122]

I tried this the other day on a souped up winuae config. And I got it to show the moving image with no background images and no grid at the start... I'll take a screenshot and post it tomorrow...

Can mk2 really be a possibly as a game on any form of Amiga?

[jotd]

we are already blessed by Richard with good OCS ports.

Specially MK is very playable! I always wanted to master MK2 but it seems that the game hates me (it also "cheats"). Too many characters & moves to learn. MK 1 is much simpler, uglier, and just pure fun, one or 2 players.

As cool as those tech tests can be, we've seen in the past (Final Fight, Street Fighter 2, ...) that it almost never ends up as a full playable game.

About the amiga source, I seem to remember that Richard could not share it, as it contains arcade source code.

[sokolovic]

Originally Posted by ransom1122

Can mk2 really be a possibly as a game on any form of Amiga?

What do you mean ? MK2 already existe on the Amiga and it is very playable.

[ransom1122]

Originally Posted by sokolovic

What do you mean ? MK2 already existe on the Amiga and it is very playable.

Arcade port

[roondar]

Originally Posted by jotd

There's another possibility: superimpose 4 HW sprites to get 12 independent colors per character (more with copper dynamic color change), but the size will be 64 max unless you reserve 3 colors for the extreme left/right of the sprites, in which case you could go wider... or complement wide moves with blitter.

I've done such a thing on Xevious port: the big ship is 96 pixels wide. I've used 4 sprites which could display 6 different colors. The original ship uses 7 colors, I removed a shade somewhere and it's almost the same. And it allowed to use the 4 other sprites with completely independent colors (player ship also used 2 sprites, plus other specially colored enemies).

MK source code (amiga or original) would be great news. I can re-do things without the source but it takes a lot more time.

That is quite a neat idea, I forgot that this makes sense to do on AGA due to the extra colours you get. Whether or not it's useful for MK2 AGA or not, it's still a useful trick to have regardless!

[jotd]

Great that noone else thought about it and now it can be attempted.

I quickly checked MK arcade. Richard Costello talked about 68020 code but he may be completely wrong as the processor is one of a TMS (Texas Instruments) series.

Also, the tiles/sprites aren't rippable from MAME menu (bad news but understandable). So maybe there's another version which is 68020 (there's a Sega Genesis version) which could be adapted, because it's already great, and possible scaled down enough so the work doesn't have to be re-done. I'm not a specialist in ripping Sega Genesis titles, sadly. Maybe the version runs on MAME and everything can be ripped like if it's a regular arcade game now that MAME and MESS have merged?

I also don't understand Richard saying that A1200 didn't exist when MK amiga was done. I remember buying MK in spring 1994 (it was coded right before that), and surely the A1200 existed. MK arcade was out in 1992.

EDIT: just tested the Genesis version on MAME. It looks so much like the amiga version... Not worth using this as a base conversion. Surely the code could be reused, but not the sprites. And the sprites are probably 16x16 blocks as Richard hinted which means that it's going to be tough to rip them properly...

Sorry for derailing the thread... Maybe another thread should be opened, and some posts migrated there... Dunno. Keeping quiet now.

[samo79]

For an eventual version for AGA better taking the PC version, or even directly the Arcade one

[jotd]

the problem with remaking such a game (MK or MK2, so I'm on topic) is that even with sprite rips (https://www.spriters-resource.com/ar.../sheet/166919/) you don't really have the "source" of the sprites, but the "solved" frames. For memory reasons, it's possible that characters are split into zones (for instance when Johnny puts his glasses on, the feet & legs are exactly in the same position)

You can't just put all frames in the memory without some sort of compression or tiling, like McGeezer did for Devil's temple, or Richard Costello told he did for the amiga version (or the megadrive assets). It would just take too much memory (MK uncompressed arcade ROM is 7MB)
So maybe arcade doesn't "tile" the sprites but for a home version (except for Neo Geo) you have to.

[dlfrsilver]

Originally Posted by jotd

the problem with remaking such a game (MK or MK2, so I'm on topic) is that even with sprite rips (https://www.spriters-resource.com/ar.../sheet/166919/) you don't really have the "source" of the sprites, but the "solved" frames. For memory reasons, it's possible that characters are split into zones (for instance when Johnny puts his glasses on, the feet & legs are exactly in the same position)

You can't just put all frames in the memory without some sort of compression or tiling, like McGeezer did for Devil's temple, or Richard Costello told he did for the amiga version (or the megadrive assets). It would just take too much memory (MK uncompressed arcade ROM is 7MB)
So maybe arcade doesn't "tile" the sprites but for a home version (except for Neo Geo) you have to.

MK sprites in arcade are stored in Bitmap Chunky mode (PC chunky). that's why mame doesn't display the graphics.

[lionagony]

Originally Posted by jotd

You're right! I just could not help commenting. I don't know any action games using 256 colors.

buzzybee mentioned Reshoot Proxima III used 256 colors, from https://eab.abime.net/showpost.php?p...&postcount=447 he says:

Stage 1 uses dual playfield, Stage 2-5 use a conventional bitmap setup with 7 or 8 bitplanes

In case of 8 bitplane setup, 224 colors are used for bitplanes, remaining 32 for sprites

As mentioned Turrican 2 AGA also used 256 colors and Ultra Violent Worlds, but those both required Fast RAM.

[lmimmfn]

Originally Posted by jotd

they say they are using 4 hardware sprites per fighter, so basically 16 specific colors per fighter for a possible width of 128 bits. It makes such a game completely possible, but the devil is in creating a playable game, with A.I, etc...

I think my calculations are exact? 4 HW sprites per character: 2 attached sprites 16 colors = 128 possible width => more than enough.


I was thinking of trying to reverse MK Amiga for AGA but what Richard Costello wrote daunted me (plus the fact that there's no source available, either Amiga or arcade)

He used bitplane tricks to display characters and said that it would be very hard to adapt the code to increase the number of bitplanes. But with hardware sprites it would be different.

Are you sure the MK sources aren't available? I downloaded them last year, they used some weird CPU but all the loops/AI routines were there and Boon listed as the author of the sources.
I can see if I can upload them somewhere if you can't find them.

Would be mega if you could transcode them but again it's a weird CPU.

[jotd]

Originally Posted by lionagony

buzzybee mentioned Reshoot Proxima III used 256 colors, from https://eab.abime.net/showpost.php?p...&postcount=447 he says:

Stage 1 uses dual playfield, Stage 2-5 use a conventional bitmap setup with 7 or 8 bitplanes

In case of 8 bitplane setup, 224 colors are used for bitplanes, remaining 32 for sprites

As mentioned Turrican 2 AGA also used 256 colors and Ultra Violent Worlds, but those both required Fast RAM.

Yes, some rare and recent games can use it. With fastmem. 128 colors I can see the point (Supercars 2 AGA uses 128 colors and is playable). 256 + sprites it means that you have to share palette with sprites (so sprites don't add any color, just speed) and I always found this annoying. But what is doable is to use the 8 first colors of the sprites for the bitmap palette, as they are not used by sprites (transparent).


So 256-32+8 = exclusive 232 colors for bitmaps. The trick also applies for ECS games with 32 colors.

Originally Posted by lmimmfn

Are you sure the MK sources aren't available? I downloaded them last year, they used some weird CPU but all the loops/AI routines were there and Boon listed as the author of the sources.
I can see if I can upload them somewhere if you can't find them.

Would be mega if you could transcode them but again it's a weird CPU.

Yes please share. Isn't this game in C? Wierd that they chose this TMS crap and the game could be remade on SNES afterwards... Just for fun I downloaded warriors shrine background, counted colors: 373 colors! ouch! downgraded to 16 colors and didn't see any difference

[str0m]

Quite an interesting 'CPU' that TMS one, only 6.25mhz apparently as well, used in quite a few games

https://en.wikipedia.org/wiki/TMS34010
https://www.system16.com/hardware.php?id=610
https://www.system16.com/hardware.php?id=611

[jotd]

I have source code for NARC. This TMS chip has a lot of similarities with 68000 after all

WAIT_FOR_Y
         SLEEP   1
         MOVE    *A13(SPR_Y_SCROLL),A14,L    ;GET CURRENT SCROLLING COORDINATE
         MOVE    @Y_CORR,A10,L                ;GET THE ONE INDICATED BY IRQ

         SUB     A10,A14                   ;THIS IS NEGATIVE OF THE OFFSET
         NEG     A14                       ;TURN IT POSITIVE (OUR OFFSET)

         MOVE    A10,*A13(SPR_Y_SCROLL),L  ;STORE THIS BACK AS OUR DISTANCE

         MOVE    *A13(SPR_Y),A8,L          ;GET Y FOR CURRENT BLOB
         ADD     A14,A8                    ;ADD CORRECTION
         MOVE    A8,*A13(SPR_Y),L          ;STORE IT BACK

         CMPI    Y_TO_DRAW*STRUCT_Y_UNIT,A8    ;ARE WE HIGH ENOUGH
         JRHI    WAIT_FOR_Y                    ;not yet....wait some more

         RETP

Axx are the 15 registers. We're talking about general purpose address/data registers like the powerpc architecture rxx registers. Transcoding would be possible with a lot of EXG instructions to use data or address registers

Fun fact:

The Amiga A2410 graphics card uses the TMS34010 and was sold in Commodore Amiga UNIX workstations, the Amiga 2500UX and 3000UX.[14] It was developed in conjunction with the University of Lowell. When running Amiga UNIX, the card supports the X Window System and gives a high resolution 8-bit display. The card can also be used when running Amiga OS, with support libraries and some Retargetable Graphics implementations.

[Amigajay]

Originally Posted by jotd

EDIT: just tested the Genesis version on MAME. It looks so much like the amiga version... Not worth using this as a base conversion.

No surprise as Probe software did most of the Mega Drive and Amiga versions, with the latter based on the MD code/graphics (MK, Alien3, T2Arcade)

[Old_Bob]

Originally Posted by lmimmfn

Are you sure the MK sources aren't available? I downloaded them last year, they used some weird CPU but all the loops/AI routines were there and Boon listed as the author of the sources.
I can see if I can upload them somewhere if you can't find them.

Would be mega if you could transcode them but again it's a weird CPU.

There are source files available for Playstation and Windows 95 ports of MK3. I can stick the zips in The Zone. There's a sub folder in the Windows source zip file with a load of TMS34010 asm files that, I guess, is the code for the coin-op that the Windows game was derived from. Along with the port's own C code and some x86 asm stuff.

I've no idea how complete either collection is, though.

Did anybody here manage to grab the MK2 stuff before it got pulled down?

B