[WIP] Tetris conversion

[jotd]

Originally Posted by ransom1122

Nice, [ Show youtube player ]

I know its super early though what i saw

1. Winuae - Full Screen, starting Medium game crashes
2. Crashed after Round 2 playing in Full Screen Window
3. Random other crashes on either window or fullscreen

Keep up the good work jotd

thanks. Can you explicit the crashes? and if it crashes because of fullscreen/window then it's not really a problem with the game.

[zzbylu]

Gameplay: [ Show youtube player ]

[Don_Adan]

Originally Posted by jotd

thanks. Can you explicit the crashes? and if it crashes because of fullscreen/window then it's not really a problem with the game.

At end of video, You can see WHDload crash.
Too bad that most players dont created Coredump reports (often only text report is enough) when tested game crashes.
For me fullscreen/window crash can be related to game too, especially if game trashes memory or registers, but maybe Im wrong.

[tomcat666]

VERY nice Loving it. Perfect really.

One question: How about 2nd button to rotate anti-clockwise ?

[malko]

It's funny as the initials I used to use to put my name in the highscore table while playing on arcade halls are present in this conversion

[jotd]

Originally Posted by tomcat666

VERY nice Loving it. Perfect really.

One question: How about 2nd button to rotate anti-clockwise ?

Good idea, but that is not that easy to code because of the special corner case of shape against the wall. Nice suggestion though.

About the crashes I think I may have an idea, stay tuned

[btp]

An old favorite of my new! I've spent quite some quarters in that one!
From the videos I watched you've made some great progress already! Simply amazing!

[DanyPPC]

Here a GlowIcon:

[jotd]

thanks for the icon!

Originally Posted by redblade

What's the code size compared to the 6502? I know that the 68k has more registers available but the instruction size jump to 16bits, and then a possible 32bits for memory location is a big step up from the 8bit.

Is 68k 4 times bigger? (excluding sound code)

Original game ROM is around 64k (not counting graphics). There are some data, at least 16k, and possibly more. I'd say that the 6502 code is around 40k, the converted assembled 68k object file is 120k. A better conversion could lead to smaller code, but there are a lot of emulation-like code to avoid rewriting everything manually.

Originally Posted by ransom1122

Nice, [ Show youtube player ]

I know its super early though what i saw

1. Winuae - Full Screen, starting Medium game crashes
2. Crashed after Round 2 playing in Full Screen Window
3. Random other crashes on either window or fullscreen

Keep up the good work jotd

creating a coredump & zoning it would be much appreciated! I see where the game crashes and that's because of a write in a non-writable memory zone, so programming/conversion error.

[ransom1122]

I didn't think it was something you were after seeing its so early alpha.. I'll see what I can do when I play test it again

[jotd]

actually the game is pretty complete except for sound and a few extra features like highscore save & better pokey random.

[ransom1122]

Originally Posted by jotd

creating a coredump & zoning it would be much appreciated! I see where the game crashes and that's because of a write in a non-writable memory zone, so programming/conversion error.

Done.. In the zone

[Amiga1992]

You've made my dream come true, JOTD

[redblade]

Originally Posted by jotd

Original game ROM is around 64k (not counting graphics). There are some data, at least 16k, and possibly more. I'd say that the 6502 code is around 40k, the converted assembled 68k object file is 120k. A better conversion could lead to smaller code, but there are a lot of emulation-like code to avoid rewriting everything manually.

Thanks , just over 3x times for the code doesn't seem too bad, I knew the Original MAC was stuffed with just 128K RAM but that is crazy.

So you just go off the MAME header find the code ROM and disassemble it and make sure the RAM Addresses which the ROM tries to access is mapped correctly? Is this where WHD comes in with mapping the low RAM address space?

[jotd]

no, it's not like that. I'm managing the real 64 memory like an emulator would do, with mapping of various regions. The game runs on 68020 without whdload and without using real zero page memory (which would cause other issues). I know some Z80 to 68000 converters which seem to do that to remain in the first 32k or 64k but not here.

Depending on the address (ram, rom, bank, eeprom) the program adds the proper offset. It can detect wrong addresses too (video & palette must be specially processed so writes have a direct effect). Game also uses a slapstic AKA bank selector, so the address converter needs to know which bank is active.

I'm really amazed that it runs great with so little reverse engineering by my end.

* < A0: offset
* flags preserved
get_address:
	PUSH_SR
	cmp.w	#0x1000,a0
	bcc.b	0f
	* RAM
	add.l	#game_ram,a0
	POP_SR
	rts
0:
	* above 0x1000
	cmp.w	#0x2100,a0
	bcc.b	0f
	lea		address_1000_20FF_loaded_message(pc),a1
	jbsr	osd_break
	POP_SR
	illegal
0:
	* above 0x2100
	cmp.w	#0x2400,a0
	bcc.b	0f
	lea		address_2100_23FF_loaded_message(pc),a1
	jbsr	osd_break
	POP_SR
	illegal
0:
	* above 0x2400
	cmp.w	#0x2600,a0
	bcc.b	0f
	* EEPROM
	lea		eeprom_2400-0x2400,a0
	POP_SR
	rts
0:
	* above 0x2600
	cmp.w	#0x4000,a0
	bcc.b	0f
	lea		address_2600_3FFF_loaded_message(pc),a1
	jbsr	osd_break
	POP_SR
	illegal	
	illegal	
0:
	* above 0x4000: caution: compare with long!!
	cmp.l	#0x8000,a0
	bcc.b	1f
	* banked ROM, check which bank
	lea		bank1_4000-0x4000,a1
	tst.b	current_bank
	bne.b	0f
	lea		bank0_4000-0x4000,a1
0:
	add.l	a1,a0
	POP_SR
	rts
1:
	* above 0x8000: ROM
	add.l	#rom_8000-0X8000,a0
	POP_SR
	rts

now to load

	GET_ADDRESS	0x00ca                     	| [$401e: lda	$00ca]
	move.b	(a0),d0                         	| [$401e: lda	$00ca]

with GET_ADDRESS as a macro

	.macro GET_ADDRESS	offset
	lea	\offset,a0
	jbsr	get_address
	.endm

This is the only sane way to do with a 6502 which doesn't have address pointers but indirect accesses from zero page. On Z80, the code was much more 68000-like, and I was using real variables. Pengo didn't even have a copy of ROM, except for the tiles of the title screen. Everything else had been converted to tables, and it could have passed for native 68000 code. Here it's not possible without understanding & rewriting everything.

Music coding is currently giving me headaches, but I'll sort it out.

[Don_Adan]

I think that easiest (maybe fastest too) way to access all regions can be something like next code:

 lea MemAreas,A0
 moveq #bank,D0
 swap D0
 move.w #offset,D0
 lea (A0,D0.L),A1

 MemAreas
 ds.b $10000*Banks

[JoeJoe]

It's impressive how quickly you implemented this port.

In my current port I use the following code:

  move.b    $30B+GRD_StandardData_8000(a5),d0    ; [$0293: ld   a,($830B)]
  lea       $314+GRD_StandardData_8000(a5),a0    ; [$0296: ld   hl,$8314]
  addq.b    #1,(a0)                              ; [$0299: inc  (hl)]
  sub.b     (a0),d0                              ; [$029a: sub  (hl)]

But even with Z80 there is a lot of mess to use a ZeroPage. And then manipulate the stack to land on the right address too.

[TjLaZer]

Here is an icon I made for Tetris.

[jotd]

Originally Posted by JoeJoe

It's impressive how quickly you implemented this port.

In my current port I use the following code:

  move.b    $30B+GRD_StandardData_8000(a5),d0    ; [$0293: ld   a,($830B)]
  lea       $314+GRD_StandardData_8000(a5),a0    ; [$0296: ld   hl,$8314]
  addq.b    #1,(a0)                              ; [$0299: inc  (hl)]
  sub.b     (a0),d0                              ; [$029a: sub  (hl)]

But even with Z80 there is a lot of mess to use a ZeroPage. And then manipulate the stack to land on the right address too.

Z80 code I've seen has a lot of stack pops, which is difficult to find and crashes.

There's also the play with H and L separately, then together, (with sometimes memory aligned on $100). Plus the infamous half carry flag (that I never saw in action yet). Depends on the programming style I guess. But 6502 is clearly more difficult to transcode into some readable code.

[PascalDe73]

Originally Posted by TjLaZer

Here is an icon I made for Tetris.

Hi,

Here My GlowIcon too (Workbench 3.5 & More)

The arcade Flyer too.

@JOTD

Nice Game Port... Congratulations

Best Regards,

Pascal