Coder shows possible way in which some elements of Street Fighter 2 might be possible - Page 16

DanScott · 04 February 2021, 17:44

Quote:

Originally Posted by Steril707

With a static image, Copper reloading can really wield amazing results even with a few colours.
Would love to see how good this works with the whole stage and not a part of it, though..

Tweaking the stages to look good with this 8+7 setup is surely a fun thing.

add/edit:

Put this together for shits and giggles.
Dans Stage with Master484s Bobs...

Looks crazy good, imo..

Now imagine this with the perspective warped floor animating, the lantern parallaxing, and chars moving and animating in 50 FPS...
And like I said, I think this is very very doable on a stock 512/512 Amiga 500.

Forgot to comment this

I tried this last night, but I think I used the wrong scale, so the characters looked too small

Yeah, it looks pretty awesome!!! Remove the orange lantern (to give more colours for the background, and confuse the conversion tool a lot less

) and put the lantern overlaid as hardware sprites, and it will look even better!! especially when moving

Lantern could use all sprites attached... or could even just use 4 x 4 colour sprites (and use the other sprites for that purple door thing on the left perhaps)

DanScott · 04 February 2021, 17:51

btw, if screen size was reduced to 288 pixels wide, I believe there could be 19 or 20 hardwarre register changes per line achievable.

You'll need 7 or 8 for the palette per line... 1 for scroll register, 1 for modulo.... hmm.. could even then have enough to do palette per line for the hardware sprites

d4rk3lf · 04 February 2021, 18:32

Using Master484 image, I've tried to do it in Photoshop.
8 colors (7 + that green transparent).

When I scrambled through the options I noticed, particularly difficult was Cammy and Ryu.. they just love to go almost desaturated bland colors. I guess other have more contrast, or more saturation in original. Maybe modfying these 2 in original, and then converting could help.

saimon69 · 04 February 2021, 19:18

The 7 color palette above ASSUME we want to keep same colors for all player fights; wonder if we can calculate custom palettes for each match and assign those there so to increase fidelity [and btw that was already mentioned in the video, albeit for 16 color palette]

[sciencefiction]
i might even go further and assume we could control copper to color each fighter differently - there will be clash spectrum style when they overlay due to the 8 pixel copper step,though;
I did propose that for Street of Rage ECS and was told not enough DMA; however here we have two fighters only (but a copperized background)
[/sciencefiction]

mcgeezer · 04 February 2021, 19:20

So I had a bit play around with some stuff.

I converted one of the frames of Ryu into a 80x128 matrix and pumped it through promotion into a tile map, you'll see in the screen shot it took up 17 tiles.

I then wrote a bit of code that would be able to flip the tiles left or right and estimate pumping them into hardware sprites.

This takes up around 62 scan lines with only chip ram enabled on stock A1200 config. So probably not the place to dump code but it seems anything goes in this thread:

Code:

ENTRY:		

	lea	$dff000,a5
	move.w	#$7fff,DMACON(a5)
	move.w	#$7fff,INTREQ(a5)
	move.w	#$7fff,INTENA(a5)

	move.w	#0,FMODE(a5)
	
	move.w	#8380,DMACON(a5)		; BPLEN+COPEN only 

.f:	COL0_BLACK

.t:	move.l	$dff004,d0
	and.l	#$1ff00,d0
	cmp.l	#50<<8,d0
	bne.b	.t

	COL0_PURPLE
	lea	SOURCE,a0
	lea	DEST,a1
	lea	TILEDATA,a2
	lea	SPRITEBUF,a3
	
	moveq	#31,d0 ; 32 tiles
	
; reverse 64 words = 1x16x16 4 bitplane tile into chip ram buffer
.lp:	rept	8
	movem.l	(a2)+,d4-d7		
	
	move.w	(a0,d4*2),(a3)
	swap	d4
	move.w	(a0,d4*2),2(a3)
	
	move.w	(a0,d5*2),4(a3)
	swap	d5
	move.w	(a0,d5*2),6(a3)

	move.w	(a0,d6*2),8(a3)
	swap	d5
	move.w	(a0,d6*2),10(a3)
	
	move.w	(a0,d7*2),12(a3)
	swap	d5
	move.w	(a0,d7*2),14(a3)
	endr
	
	dbf	d0,.lp
		
	COL0_BLACK
	bra	.f

Likely would work for two player and plenty of optimisations could be made.

Big big job though to convert all them frames.

tarr · 04 February 2021, 19:22

Quote:

Originally Posted by Master484

I tested my palette to arcade characters too:

So indeed it would seem that 8 colors is enough for this game.

Sorry, but judging the results it appears that 8 color are NOT ENOUGH AT ALLL for this game!

saimo · 04 February 2021, 19:25

Quote:

Originally Posted by mcgeezer

Code:

ENTRY:        

    lea    $dff000,a5
    move.w    #$7fff,DMACON(a5)
    move.w    #$7fff,INTREQ(a5)
    move.w    #$7fff,INTENA(a5)

    move.w    #0,FMODE(a5)
    
    move.w    #8380,DMACON(a5)        ; BPLEN+COPEN only 

.f:    COL0_BLACK

.t:    move.l    $dff004,d0
    and.l    #$1ff00,d0
    cmp.l    #50<<8,d0
    bne.b    .t

    COL0_PURPLE
    lea    SOURCE,a0
    lea    DEST,a1
    lea    TILEDATA,a2
    lea    SPRITEBUF,a3
    
    moveq    #31,d0 ; 32 tiles
    
; reverse 64 words = 1x16x16 4 bitplane tile into chip ram buffer
.lp:    rept    8
    movem.l    (a2)+,d4-d7        
    
    move.w    (a0,d4*2),(a3)
    swap    d4
    move.w    (a0,d4*2),2(a3)
    
    move.w    (a0,d5*2),4(a3)
    swap    d5
    move.w    (a0,d5*2),6(a3)

    move.w    (a0,d6*2),8(a3)
    swap    d5
    move.w    (a0,d6*2),10(a3)
    
    move.w    (a0,d7*2),12(a3)
    swap    d5
    move.w    (a0,d7*2),14(a3)
    endr
    
    dbf    d0,.lp
        
    COL0_BLACK
    bra    .f

I guess those are meant to be (a3)+?

mcgeezer · 04 February 2021, 19:30

Quote:

Originally Posted by saimo

I guess those are meant to be (a3)+?

No, i did that on purpose to increase the cycle count as might not have the luxury of using post increment addressing.

saimo · 04 February 2021, 19:42

Quote:

Originally Posted by mcgeezer

No, i did that on purpose to increase the cycle count as might not have the luxury of using post increment addressing.

I'm not sure I get it but, even then, wouldn't an adda.w #16,a3 (or adda.l Rx,a3, if you have a spare register) be needed at the end of the block?

saimon69 · 04 February 2021, 19:43

Quote:

Originally Posted by tarr

Sorry, but judging the results it appears that 8 color are NOT ENOUGH AT ALLL for this game!

Is depending of what you are looking for - for OCS/ECS better than this can hardly go [to keep fluidity] - for AGA you might have a point

If you are looking at arcade looking SF2 in A500 that runs at 50 FPS come back in the future with 20/30 more thread pages and MAYBE some of the hurdles will be solved :P

mcgeezer · 04 February 2021, 19:48

Quote:

Originally Posted by saimo

I'm not sure I get it but, even then, wouldn't an adda.w #16,a3 (or adda.l Rx,a3, if you have a spare register) be needed at the end of the block?

Tomorrow i’ll code it properly and see what the actual metrics are, the point of this code was simply to provide a ball park figure to understand if it is worth carrying on.

I think it’s worth looking at the next step which is a proof of concept code.

DanScott · 04 February 2021, 19:52

Quote:

Originally Posted by mcgeezer

Tomorrow i’ll code it properly and see what the actual metrics are, the point of this code was simply to provide a ball park figure to understand if it is worth carrying on.

I think it’s worth looking at the next step which is a proof of concept code.

don't forget the 50/50 flip

Tigerskunk · 04 February 2021, 20:40

Quote:

Originally Posted by tarr

Sorry, but judging the results it appears that 8 color are NOT ENOUGH AT ALLL for this game!

I politely disagree here..

vulture · 04 February 2021, 20:43

I find d4rk3lf's modified palette more suitable despite the lack of contrast compared to Master484's original

saimon69 · 04 February 2021, 20:50

Quote:

Originally Posted by Valken

I agree it seems the direction has strayed from how to make a near arcade perfect version but ends up setting the bar at an Amiga 500 with 1 MB of ram. AGAIN. Sigh...

However, you solve the small machine problems and you solved it for all; just scale up

Tigerskunk · 04 February 2021, 20:55

With the arcade sprites the scene looks even more amazing.

Mirror matches will be a problem though with this palette. But they would be anyway on this setup.

mcgeezer · 04 February 2021, 20:55

Same as what happened in the epic Rygar thread trying to port it to OCS.

Won’t happen. AGA.... maybe

jotd · 04 February 2021, 21:00

Just make it AGA but with a limited set of colors to avoid wasting too much blitter bandwidth

Wanting to go all ECS is going to be painful and will look very much like the US Gold version in the end... Rygar was made in 1986 and was hard enough with AGA. Steet Fighter 2 was made in 1993 on a much superior hardware. Let's be realistic

A lot of amiga users have a A1200. Others are stuck with thousands of old excellent games.

If a AGA version is made, maybe it's possible to downgrade it to ECS with some extra effort.

saimon69 · 04 February 2021, 21:02

[sciencefiction]
What is the player main color is transparent and the game plots a one bitplane mask with the mirror player color on BEFORE the player bob? Those will be two plots but somehow solve the different color problem
[/sciencefiction]

[sciencefiction2]
What is the player main color is transparent and the game plots a one bitplane part with the mirror player color on AFTER the player bob? Those will be two plots but somehow solve the different color problem
[/sciencefiction2]

btw those are the two ways that in my code illitterate analysis would explain shadow fighter mirror player different color - a third one could be plot player 1 - change palette - plot player 2 - next raster row plot player1 - change palette - plot player 2 etc...

Tigerskunk · 04 February 2021, 21:06

Quote:

Originally Posted by vulture

I find d4rk3lf's modified palette more suitable despite the lack of contrast compared to Master484's original

I tried to do a mock up with that like I did with Master484's sheet above, but it really didn't jive with the colourful looking background, imo.

04 February 2021, 18:32	#303
d4rk3lf Registered User Join Date: Jul 2015 Location: Novi Sad, Serbia Posts: 1,685	Using Master484 image, I've tried to do it in Photoshop. 8 colors (7 + that green transparent). When I scrambled through the options I noticed, particularly difficult was Cammy and Ryu.. they just love to go almost desaturated bland colors. I guess other have more contrast, or more saturation in original. Maybe modfying these 2 in original, and then converting could help. Attached Thumbnails

04 February 2021, 19:18	#304
saimon69 J.M.D - Bedroom Musician Join Date: Apr 2014 Location: los angeles,ca Posts: 3,590	The 7 color palette above ASSUME we want to keep same colors for all player fights; wonder if we can calculate custom palettes for each match and assign those there so to increase fidelity [and btw that was already mentioned in the video, albeit for 16 color palette] [sciencefiction] i might even go further and assume we could control copper to color each fighter differently - there will be clash spectrum style when they overlay due to the 8 pixel copper step,though; I did propose that for Street of Rage ECS and was told not enough DMA; however here we have two fighters only (but a copperized background) [/sciencefiction] Last edited by saimon69; 04 February 2021 at 19:46.

04 February 2021, 19:20	#305
mcgeezer Registered User Join Date: Oct 2017 Location: Sunderland, England Posts: 2,702	So I had a bit play around with some stuff. I converted one of the frames of Ryu into a 80x128 matrix and pumped it through promotion into a tile map, you'll see in the screen shot it took up 17 tiles. I then wrote a bit of code that would be able to flip the tiles left or right and estimate pumping them into hardware sprites. This takes up around 62 scan lines with only chip ram enabled on stock A1200 config. So probably not the place to dump code but it seems anything goes in this thread: Code: ENTRY: lea $dff000,a5 move.w #$7fff,DMACON(a5) move.w #$7fff,INTREQ(a5) move.w #$7fff,INTENA(a5) move.w #0,FMODE(a5) move.w #8380,DMACON(a5) ; BPLEN+COPEN only .f: COL0_BLACK .t: move.l $dff004,d0 and.l #$1ff00,d0 cmp.l #50<<8,d0 bne.b .t COL0_PURPLE lea SOURCE,a0 lea DEST,a1 lea TILEDATA,a2 lea SPRITEBUF,a3 moveq #31,d0 ; 32 tiles ; reverse 64 words = 1x16x16 4 bitplane tile into chip ram buffer .lp: rept 8 movem.l (a2)+,d4-d7 move.w (a0,d42),(a3) swap d4 move.w (a0,d42),2(a3) move.w (a0,d52),4(a3) swap d5 move.w (a0,d52),6(a3) move.w (a0,d62),8(a3) swap d5 move.w (a0,d62),10(a3) move.w (a0,d72),12(a3) swap d5 move.w (a0,d72),14(a3) endr dbf d0,.lp COL0_BLACK bra .f Likely would work for two player and plenty of optimisations could be made. Big big job though to convert all them frames. Attached Thumbnails

04 February 2021, 20:55	#316
Tigerskunk Inviyya Dude! Join Date: Sep 2016 Location: Amiga Island Posts: 2,793	With the arcade sprites the scene looks even more amazing. Mirror matches will be a problem though with this palette. But they would be anyway on this setup. Attached Thumbnails

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Street Fighter 2	weirdreams	Retrogaming General Discussion	4	20 June 2012 23:15
Super Street Fighter 2	Retro1234	project.Sprites	94	12 December 2008 11:20
street fighter	stuntpup	project.WHDLoad	5	30 August 2007 20:45
Street Fighter III	Muzkat	Retrogaming General Discussion	11	14 August 2007 00:55
[Fixed] Street Fighter II	Amigaboy	HOL data problems	5	30 December 2002 21:34

04 February 2021, 17:51	#302
DanScott Lemon. / Core Design Join Date: Mar 2016 Location: Tier 5 Posts: 1,213	btw, if screen size was reduced to 288 pixels wide, I believe there could be 19 or 20 hardwarre register changes per line achievable. You'll need 7 or 8 for the palette per line... 1 for scroll register, 1 for modulo.... hmm.. could even then have enough to do palette per line for the hardware sprites

04 February 2021, 20:43	#314
vulture Registered User Join Date: Oct 2007 Location: Athens , Greece Posts: 1,856	I find d4rk3lf's modified palette more suitable despite the lack of contrast compared to Master484's original

04 February 2021, 20:55	#317
mcgeezer Registered User Join Date: Oct 2017 Location: Sunderland, England Posts: 2,702	Same as what happened in the epic Rygar thread trying to port it to OCS. Won’t happen. AGA.... maybe

04 February 2021, 21:00	#318
jotd This cat is no more Join Date: Dec 2004 Location: FRANCE Age: 52 Posts: 8,360	Just make it AGA but with a limited set of colors to avoid wasting too much blitter bandwidth Wanting to go all ECS is going to be painful and will look very much like the US Gold version in the end... Rygar was made in 1986 and was hard enough with AGA. Steet Fighter 2 was made in 1993 on a much superior hardware. Let's be realistic A lot of amiga users have a A1200. Others are stuck with thousands of old excellent games. If a AGA version is made, maybe it's possible to downgrade it to ECS with some extra effort.

04 February 2021, 21:02	#319
saimon69 J.M.D - Bedroom Musician Join Date: Apr 2014 Location: los angeles,ca Posts: 3,590	[sciencefiction] What is the player main color is transparent and the game plots a one bitplane mask with the mirror player color on BEFORE the player bob? Those will be two plots but somehow solve the different color problem [/sciencefiction] [sciencefiction2] What is the player main color is transparent and the game plots a one bitplane part with the mirror player color on AFTER the player bob? Those will be two plots but somehow solve the different color problem [/sciencefiction2] btw those are the two ways that in my code illitterate analysis would explain shadow fighter mirror player different color - a third one could be plot player 1 - change palette - plot player 2 - next raster row plot player1 - change palette - plot player 2 etc...

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)