When a projectile collide with a player, particles exist. Depending from the projectile speed, different particle "profile" is generated. Arcade and snes have the same "profiles" and particle code and normally would have the same result in each situation. If you can tell a difference, please tell me to correct it.

I see. It was only how I interpreted Senor Ventura's message, guess we need to wait until he clears that up

Conn wrote:I think he wants those fireworks sparkles whenever a ha do ken collides (e.g., blocked or enemy hit), and not only when two collide.

I only wanted to provide inspiration xD

gizaha wrote:When a projectile collide with a player, particles exist. Depending from the projectile speed, different particle "profile" is generated. Arcade and snes have the same "profiles" and particle code and normally would have the same result in each situation. If you can tell a difference, please tell me to correct it.

Yes, i think that the behaviour is practically the same, but the other form is a lot more spectacular.

I tested this game and it was more than good !!!
Gizaha is a monster !!!
Congratulations boy and thanks for everything !!!

Now i got it. I agree that's more spectacular but i insist at staying faithful to the arcade and using the arcade codes.

Hows it looking on the SD2SNES? Any progress?

Fixed some errors but Sd2snes has bug in the dma from msu. Sent message to Ikari, he read it, waiting for fix. Otherwise i will need to fix THIS PROBLEM ALSO after:
*Reverse engineer snes ki.
*Reverse engineer arcade ki.
*Create gimp script for converting pictures to quantized 4bpp multipalette for fmv (the hardest of all).
*Grab sprites from arcade even the hard ones (like char select 360 frames)
*Create script for generating ki snes sprites.
*Create arcade sprite viewer (out of boredom)
*Create fmv player
All these while receiving countless rejections for help from others. I could write a whole page with their names.

We will wait. Otherwise, its not the first time i will correct another's man code.

Sounds tough. Its a shame that other skilled people don't want to help you. I've been following your work on this project for awhile now, hopefully someday you get this running on the SD2SNES running smooth and how you want. Keep up the good work tho brother.

gizaha wrote:Now i got it. I agree that's more spectacular but i insist at staying faithful to the arcade and using the arcade codes.

By the way, what about color math for these particle sprites?, Do these has some subpalette sharing with another sprites?.


edit: With the actual size of character sprites and animations, How much of bandwidth is reaching the DMA?

Señor Ventura wrote:
By the way, what about color math for these particle sprites?, Do these has some subpalette sharing with another sprites?.

They share projectile's palette. That's why Jago's orange fireball has orange particles etc. (except when they touch Sabrewulf's stage fireplace they turn red).

They don't have color math. I've made experiments, this looks better.
If they use transparency, because every particle has it's own properties, it must have it's own palette. It's getting complicated. I was planning to replace the stock particle graphics with bigger, but now i don't find the need.

Señor Ventura wrote:
With the actual size of character sprites and animations, How much of bandwidth is reaching the DMA?

Very small. Player sprite size is usually $600 (max $0c00) *2 = $c00. It can easily use full scale arcade sprites.
Character select sprites are bigger than $0c00 but they DMAing with frame difference.

gizaha wrote:

Señor Ventura wrote:With the actual size of character sprites and animations, How much of bandwidth is reaching the DMA?

Very small. Player sprite size is usually $600 (max $0c00) *2 = $c00. It can easily use full scale arcade sprites.
Character select sprites are bigger than $0c00 but they DMAing with frame difference.

I am amazed by that statement alone. It is exciting to think of a near arcade quality big sprites KI playable on the snes. I can't even play it anymore without the current patch.

On overclocked Wii U with the Snes Wii emulator, it runs very good, except the Vs FMVs, and some heavy slowdown for some seconds after the announcer says 'ready'.

@gizaha Once a time was a killer instinct beta with a lot of smooth animations, and the DMA seemed to endure that level of tile loadings.

Here is:


And this:



About the colors... there are 128 colors in 8 subpalettes of 15+1 colors, Do you think that killer instinct uses it well?... and in your opinion, Could be there room for improvement?.

Thank you for sharing your experience with the programming of this game

1) I can't see the videos (video is private). I guess you're reffering to the E3 '95 beta version with more frames.
2) (assuming that 1 is correct) You're confuzing DMA with more animation frames. DMA is the copy procedure of data to vram. It's made every frame (for one frame). If a 1 second length move animation has 8 or 60 frames, it's no difference for DMA. It's uploading player data each frame. So, you can have 60fps animations for both players and be within limit.
3) I don't think palette can be improved. First 8 for BG used fully.
The rest 8 for sprites have some static that in theory can be dynamic but i can't think a reason. Nothing is missing.

One improvement might be to dynamically refresh the player palette each frame to match the existing animation. Some specific frames will benefit from that. 15 colors are more than enough for one sprite frame but using the same 15 colors at ALL player animations can show it's weakness.

Epilogue: Don't forget than in theory snes KI can highly mimic arcade version. I'm not even talking about SA1 or FX2.

Try again, please, these two videos must can be seen now.

As you can see in the videos, there is a point that the game suffer slowdowns. I asked about the DMA cause the more you have DMA active, the more you can't dispose the cpu for its tasks.

There is enough DMA for 60 tic animations there, but maybe the cpu is drowned, cause in every frame is stopped too much time. I don't know really, so i'm glad i'm wrong xD


About the colors... Do all the eight subpalettes are used completely, then?.

Last edited by Señor Ventura on Sun 14 Jun 2020 - 6:13; edited 1 time in total

Señor Ventura wrote:there is a point that the game suffers slowdowns. I asked about the DMA cause the more you have DMA active, the more you can't dispose the cpu for its tasks.

It's not DMA problem but main code's (the exact opposite).
Main code start running at vline 0. Optimally will finish before 224 and wait for it. At 224, DMA is refreshing vram. Next frame, repeat.

Specific slowdown was caused by too many blood objects. Main code finish after 224, so DMA was run at the next frame's 224 line, thus 30fps (because main code didn't run in this frame).
In final rom they add speed tweaks. Beta roms that leaked at the internet have the tweaks also, so E3 rom was kind of Alpha (had direct arcade code without tweaks).
I've also added speed tweaks here and there.

60fps animations don't cause lag.
Subpalettes used completely.

Fixed some errors but Sd2snes has bug in the dma from msu. Sent message to Ikari, he read it, waiting for fix. Otherwise i will need to fix THIS PROBLEM ALSO after:
*Reverse engineer snes ki.
*Reverse engineer arcade ki.
*Create gimp script for converting pictures to quantized 4bpp multipalette for fmv (the hardest of all).
*Grab sprites from arcade even the hard ones (like char select 360 frames)
*Create script for generating ki snes sprites.
*Create arcade sprite viewer (out of boredom)
*Create fmv player
All these while receiving countless rejections for help from others. I could write a whole page with their names.

We will wait. Otherwise, its not the first time i will correct another's man code.

Ikari code is open source, if you are going to look into/fix it, could you maybe look also in this bug (only if you have time)?
http://krikzz.com/forum/index.php?topic=9737.0

I think Ikari forgot about to fix... I tried to look at the source but didn't find a possible bug source, it should be easy though, if you know where to look.

This is the most amazing hack I've ever seen!
I'm a fan of Killer Instinct and SNES, now it's hard to play the original version.

Once again, thank you very much for your amazing work!

Last edited by ErivandoBR on Wed 24 Jun 2020 - 9:06; edited 2 times in total (Reason for editing : Correction)

Congratulations Gizaha. One day I hope you can run on sd2snes.

Thanks, even it's not at my hand. I saw some video at youtube with extreme combos at arcade version and i want these players to test theses combos in snes version, if it's now applicable. Otherwise, i want to know the differences to correct them. I tried to contact one or two of them and for the 1000 time, no reply as usual.

Which combos? Im really good at this game and can test them out.

https://youtu.be/zCXkR4pkGfU?list=FLTHCxTuu6fMysCfCwMhN36A&t=516

I like this player's combos. At TJ combo, after lk autodouble he uses 2 manual mk. All these years i didn't knew this specific autodouble, nor the manual mk. This combo can achieved in snes now, but i don't know for the other players.

Oh yes I know this combo. On the arcade version if you do a ultimate after the 2 manual hits it will cause a knockdown and then TJ knocks him to the screen. It looks really cool. Does it react the same on the snes?

gizaha wrote:https://youtu.be/zCXkR4pkGfU?list=FLTHCxTuu6fMysCfCwMhN36A&t=516

I like this player's combos. At TJ combo, after lk autodouble he uses 2 manual mk. All these years i didn't knew this specific autodouble, nor the manual mk. This combo can achieved in snes now, but i don't know for the other players.

Hello, I'm also very good at this game!
I'll try here too!

Ran the newest version of the patch with the Wii, getting better and better!

While the VS screens won't show (too slow), the first 4 o 5 matches start immediately if you skip it, which wasn't the case before. After that, it will lag for some seconds after starting the match before you can control the character again.

Playing with Orchid: If you do an Ultra combo to Riptor in his stage, the character "floats" appearing from below the floor, and finally falls again. Doing a long combo or Ultra to Cinder slows down the game considerably (but no frameskip).

Round 2 on the Jago stage creates some lag too.

So far the most taxing fights seem to be Cinder and Glacius, the last one being the cause of extreme frame skipping, which seems to be the emulation catching up.

Runs really great on the PC, of course. Only the slowdowns against Cinder remain.

Awesome job, Gizaha!

niuus wrote:While the VS screens won't show (too slow)

Bug didn't solved, it's on workaround.

Playing with Orchid: If you do an Ultra combo to Riptor in his stage, the character "floats" appearing from below the floor

He's not floating, it's his shadow.

Doing a long combo or Ultra to Cinder slows down the game considerably (but no frameskip)

Game is not programmed to have frameskip.

Round 2 on the Jago stage creates some lag too.

Didn't knew that, i don't have a wii. I only test it at bsnes and sd2snes.

So far the most taxing fights seem to be Cinder and Glacius, the last one being the cause of extreme frame skipping, which seems to be the emulation catching up.

Because Glacius has 3 cloud of mist around him, Cinder has flame particles.
3 clouds, 8 blood objects, 1 star object, 2 player objects, hitbox detection and all of these have gravity and ground detection, toghether with a small lag from spc each time it's called to play a hit sfx.

Awesome job, Gizaha!

Thanks.