[ASM68000] rubik's

[Arnaud.68000]

Hi,


30 years after, i worked for monthes to code a 3D rubik's cube... with rotations of faces.


Here is the picture of the result :







Please find the code there (Sorry all my comments are in french).



https://github.com/Amiga68000/Rubiks...tag/2021-02-08


Don't hesitate to give me feedback if you find bug, optimizations or have suggestions or ideas.

Any question, please ask.

If i find motivation, i will try to implement filled shapes.



Thanks & have fun


Arnaud

[a/b]

Did one of those back in 92-93 with filled vectors and white line vectors for the grid.
I used 2x3 dual playfield so I only had to deal with 2 easy peasy convex objects, each in its own playfield, rotated two sets of 8 corner points and the rest was generated with unit vectors.

[Valken]

Very cool! You should make it into a full demo that we can play around with or randomize the rotations then try to solve it automatically!

[heavy]

warning : somes bsr (pc) and bra (pc). work with jsr not bsr. doesn't work with asmone

Originally Posted by a/b

Did one of those back in 92-93 with filled vectors and white line vectors for the grid.
I used 2x3 dual playfield so I only had to deal with 2 easy peasy convex objects, each in its own playfield, rotated two sets of 8 corner points and the rest was generated with unit vectors.

I have seen only two rubik on A500. one from French group Intryx and one in the trackmo of Binary. with white line vectors for the grid : a great rubik!
is it this one ?

[a/b]

Yep, the latter.

[heavy]

Made In Croatia / Binary (1993) :
http://janeway.exotica.org.uk/release.php?id=9158

impressive rubik's cube
can you tell more about how that work ? (differences with the one by Arnaud)
thanks

[a/b]

Eh, it's simple stuff made for a demo, nothing more nothing less. Didn't check Arnaud's code in detail, french gives me exception #4 ;p, but it looks like a completely different approach.
As I said, dual playfield is the critical part, makes rendering trivial. Simple "script", takes a series of +/- 1-6 (6 possible moves, sign determines rotation direction), selects one of 6 predefined poly/corner sets, generates inner 3d points, does some math and throws everything at the screen, Tops everything off with shiny line vectors to hide ugly edges ;P. The usual convex vector demo stuff.
Playfield priority (and the consequent remapping of colors) is based on the sum of corner z coords.

[Arnaud.68000]

Why not Valken, but including a solver will be a little bit harder for me
First a simple demo should be easier.

Made In Croatia / Binary (1993)

Very impressive to calculate and display in real time.

french gives me exception #4 ;p

Nooooo ... if you want I will send my code to Buckingham Palace, I know that there several people speak French without #4 exception

@a/b, not sure to understand in details the method you describe :
If you do (or not) a rotation of a layer you use 2 playfields for 2 3D-objects, 1 for the body(without the layer) + 1 for the layer in rotation
You calculate only 8 corners per 3D-object, using a linear interpollaton, you calculate shapes and fill them.
i'm right ?
Why 2 playfiels, is it to simplify the filling ?


What i did is more complicated (i calculated and display each visible shape) and must use more CPU than your method


"Merci pour l'astuce" ;p

I will try to think how to implement it

[heavy]

I just compiled your source with asm2 and winuae 4400 : nothing on the screen, except text and debug : no cube
I see the cube with debugdisplayraster=1

[Arnaud.68000]

I don't understand,
I just download and tried the source and it works with my config :
- WinUAE 64.4.4.0 (A500 / 1Mo / true time)
- ASMTwo
What is your config under WinUae ? A500 or more ?

warning : somes bsr (pc) and bra (pc). work with jsr not bsr. doesn't work with asmone

Ok "Heavy", I will try ASMOne to clean my code

[heavy]

1 bsr and 1 bra to clean, it's enough
but on winuae32 4.4.0 (A500) : no cube on the screen
only with debugdisplay_raster=1
with ASMTwo !
and no more with asmone

[Arnaud.68000]

You remind me a bug.

line 1680 :

	dc.w	COLOR01,MIRROR_SCROLL_COLOR
	dc.w	COLOR02,MIRROR_SCROLL_COLOR
	dc.w	COLOR03,MIRROR_SCROLL_COLOR
	;dc.w	COLOR04,MIRROR_SCROLL_COLOR
	dc.w	COLOR05,MIRROR_SCROLL_COLOR
	dc.w	COLOR06,MIRROR_SCROLL_COLOR
	dc.w	COLOR07,MIRROR_SCROLL_COLOR

Try to comment all the lines.

On my machine if I uncomment COLOR04, the cube is not displayed
Maybe you need uncomment more than 1 line.

I didn't find time to investigate this bug .. if someone has an idea ...

[heavy]

nop, and now I have a guru
problem with the copperlist

[heavy]

line 1499
change

	move.l	#COPSIZE-1,d0 ; problem with COPSIZE/4. works also with COPSIZE/2 .w
	move.l	#CopListDataStart,a0
	move.l	copperlist_PTR,a1
.l2:
	move.b	(a0)+,(a1)+ ; .b or .w if copsize/2 <- .L
	dbf		d0,.l2

your code is already in chip with section code_c
you don't need to allocmem in chip and copy the coplist

[Arnaud.68000]

which machine do you emulate ?

Bellow extract of my config .uae

config_description=Environnement ASM sur DD WB1.3
config_hardware=true
config_host=true
config_version=4.4.0
...
comp_catchdetect=true
win32.rom_path=./
win32.floppy_path=./
win32.hardfile_path=./
; host-specific
win32.middle_mouse=true
win32.soundcard=0
win32.soundcardname=WASAPIefault Audio Device
; common
use_gui=yes
use_debugger=false
kickstart_rom_file=.\10 - Rom\Kickstart 1.3.rom
kickstart_rom_file_id=C4F0F55F,KS ROM v1.3 (A500,A1000,A2000)
kickstart_ext_rom_file=
flash_file=
cart_file=
rtc_file=
kickshifter=false
floppy_volume=33
...
nr_floppies=2
floppy_speed=100
cd_speed=100
...
cachesize=0
...
bsdsocket_emu=false
synchronize_clock=false
maprom=0x0
parallel_postscript_emulation=false
parallel_postscript_detection=false
ghostscript_parameters=
parallel_autoflush=5
clipboard_sharing=true
...
immediate_blits=false
waiting_blits=automatic
ntsc=false
genlock=false
chipset=ocs
chipset_refreshrate=49.920410
collision_level=playfields
chipset_compatible=A500
rtc=MSM6242B
resetwarning=false
cia_todbug=true
fastmem_size=0
debugmem_start=0x0
debugmem_size=0
mem25bit_size=0
a3000mem_size=0
mbresmem_size=0
z3mem_size=0
z3mem_start=0x40000000
bogomem_size=2
gfxcard_hardware_vblank=false
gfxcard_hardware_sprite=true
gfxcard_multithread=false
chipmem_size=1
cpu_speed=real
cpu_throttle=0.0
cpu_type=68000
cpu_model=68000
cpu_compatible=true
cpu_24bit_addressing=true
cpu_data_cache=false
cpu_multiplier=2
cpu_cycle_exact=true
cpu_memory_cycle_exact=true
blitter_cycle_exact=true
cycle_exact=true
fpu_strict=false
rtg_nocustom=true
rtg_modes=0x212
debug_mem=false
log_illegal_mem=false
kbd_lang=us
filesystem2=rw,DH0:SYS:.\15 - HardDisk\SYS_WB1.3,0
uaehf0=dir,rw,DH0:SYS:.\15 - HardDisk\SYS_WB1.3,0
filesystem2=rw,DH1:PROG:.\15 - HardDisk\Prog,-128
uaehf1=dir,rw,DH1:PROG:.\15 - HardDisk\Prog,-128
filesystem2=rw,DH2:Sources:.\15 - HardDisk\Sources,0
uaehf2=dir,rw,DH2:Sources:.\15 - HardDisk\Sources,0
...

[Arnaud.68000]

OK for the CODE_C, thanks

[Arnaud.68000]

Originally Posted by heavy

line 1499
change

	move.l	#COPSIZE-1,d0 ; problem with COPSIZE/4
	move.l	#CopListDataStart,a0
	move.l	copperlist_PTR,a1
.l2:
	move.b	(a0)+,(a1)+ ; .b <- .L
	dbf		d0,.l2

your code is already in chip with section code_c
you don't need to allocmem in chip and copy the coplist

Is "move.l" the cause of your GURU ?

[heavy]

there is an issue with your copperlist
if I put a comment somewhere, screen stay black

[a/b]

Originally Posted by Arnaud.68000

You calculate only 8 corners per 3D-object, using a linear interpollaton, you calculate shapes and fill them.
i'm right ?

Yes, 8 corner points per object and the rest is interpolated.

Originally Posted by Arnaud.68000

Why 2 playfiels, is it to simplify the filling ?

Yes, 1 blit to clear, draw all the edges in fill mode, and finally 1 blit to fill. Also, calculate min/max x/y so you don't have to clear/fill the entire buffer.

[Arnaud.68000]

Originally Posted by heavy

there is an issue with your copperlist
if I put a comment somewhere, screen stay black

I don't have this issue, but I know I have the pb if I remove a comment in the miror color table
I will investigate, maybe a pb with the VBL ?