PIXI V2 to PIXI V3

[Ramiro]

PIXIv2 > PIXI v3​

version 0.5.0 - looking for betatesters!​

 ​

Description

Do you want to have the nice things of PIXI v3 in MV. Well, after some reading of the core classes I have somehow good news for you.

It's doable, even if it's hard. I can't completely do it alone... but as a community effort I'm sure we can pull it off.

This also will aim to improve speed of core classes as we improve the pixi module.

The idea is to leave almost anything intact from the outside, any public function should remain as it, and if we need to change it for performance reasons, we should add it to a deprecated value and nothing else.

Progress

Battlebacks are shown by now, but there is a strange flickering for sprites...

So looking for that solution now...

How do I start?

First, create a new RM Project to put this files in.

Then edit as you need.

Any plugin you want to try, any thing you want to add outside of the core will be ignored, and if you somehow add a file you should add it to a .gitignore file.

You will need a git manager like source tree or githu or console.

You can check repositories here.

Even if nobody actually helps I am pretty sure I'll finish it sooner or later (hope before PIXI v4 is out...)

Bugs

Github does have an issue tracker, so you should pick from there what you need.

But for the moment, I need to change all of the classnames of the core. So waint at least for a while untill I start to get with deprectaed things and name changes...

UPDATE:

I actually need people testing this, like, a lot, so if you please, can yo test it?

License

It is going to be in the public domain, but he original code is still held by kadokawa and spike chunsoft, so even after migration you will still need a license of mv in order to use this.

 

[Dr.Yami]

Hope this will work out ~

/me cheers Ramiro

 

[tati light]

I looked at the v3 improvements and they are pretty nice. I hope this can succeed at some point.

I would love to help as well, but I'm a total newbie for this kind of code.

 

[Ramiro]

It is working better than expected, for now I'm having some minor visual issues and just problems with FPS rates...
But the rest works...

I may rewirte tilemap or some classes to be bit more lightweight, and because the standard tilemap is a huge fps drop...

 

[Kane Hart]

It is working better than expected, for now I'm having some minor visual issues and just problems with FPS rates...

But the rest works...

I may rewirte tilemap or some classes to be bit more lightweight, and because the standard tilemap is a huge fps drop...

Yeah I agree here the vanilla tilemap is just a little ridiculous. I feel like it was written once and never touched again. Good programmers rewrite things over and over till their boss says stop release date is next week! lol

 

[Tsukihime]

What's wrong with the tilemap?

 

[Kane Hart]

What's wrong with the tilemap?

I'm not much of a programmer if at all but trying to look through how the rendering and tilemaps load they don't seem to be correctly buffered if at all and it gets pretty taxing. You really notice this on mobile phones with 60 fps support but when they move around you get quite a bit of jittering. Now a really good performing phone would sure be fine more then likely but I find personally there has to be some bad optimizations when it comes to the entire tile loading system.

Maybe someone with better experience and understanding though can say something else. 

I just personally fine my experience over multiple devices to be the biggest bottleneck. 

 

[Ramiro]

They are working currently at 40 fps...

They do look good, because I could fix the rects of them...

But you have to see how broken it was two days ago...

Still I need to improve speed

 

[Hackerham]

Ok, add me in Skype, I'll help

 

[Ramiro]

Added,

lets see what we can do now

 

[Kane Hart]

I tested it it earlier today and the performance felt almost the same as vanilla so I say your doing well. 

 

[Hackerham]

Performance is awful. I looked at profile of https://gameofbombs.com VS rpgmaker pixiv2 and pixiv3 versions, and rpgmaker is eating far more resources It should be on 3-4% level, but its on 55% for me.

 

[Kane Hart]

Performance is awful. I looked at profile of https://gameofbombs.com VS rpgmaker pixiv2 and pixiv3 versions, and rpgmaker is eating far more resources It should be on 3-4% level, but its on 55% for me.

Ahh I never checked that sort of performance I only checked fps and mine was steady for both oddly enough.

 

[Ramiro]

Still, if we can make improvements over the original code and mantain the api... we'll do that...

I already made some improvements here and there, but they are small

 

[Kane Hart]

Still, if we can make improvements over the original code and mantain the api... we'll do that...

I already made some improvements here and there, but they are small

Oh yeah for sure. I the tile rendering or whatever it's called improves. I noticed it tends to be one the higher bottlenecks. 

 

[Ramiro]

I could explain why:

The tilemap on mv isn't rendered as you moved, but the entire tileset is rendered each time.

So each time you have to render something you are passing 4-5 giant textures in memory wich leads to slowdown.

The idea of using tilesets is to draw only the area you see and ignore the rest. For that you would have to mantain a list of tiles (wich are small textures) and pass only the ones you will draw into memory, then batch draw them on screen...

It's a bit hard to implement, so don't expect those for now, not even by the official pixi2 version

 

[Hackerham]

Yay, tileset optimization! I love that!

 

[Kane Hart]

I could explain why:

The tilemap on mv isn't rendered as you moved, but the entire tileset is rendered each time.

So each time you have to render something you are passing 4-5 giant textures in memory wich leads to slowdown.

The idea of using tilesets is to draw only the area you see and ignore the rest. For that you would have to mantain a list of tiles (wich are small textures) and pass only the ones you will draw into memory, then batch draw them on screen...

It's a bit hard to implement, so don't expect those for now, not even by the official pixi2 version

Thanks for explaining it.

 

[Hackerham]

https://github.com/holywyvern/mv-pixi-upgrade/issues/2 - look at that!

5-10x  performance boost on tilemaps. Tested on map which have 30% animated water.

That thing doesnt depend on pixi, its just common sence: DONT DRAW TILEMAPS FROM CANVAS IF THERE IS AN IMAGE.

I use that optimization in gameofbombs.com , 1920x1080 tilemaps work like a charm.

 

[Kane Hart]

https://github.com/holywyvern/mv-pixi-upgrade/issues/2 - look at that!

5-10x  performance boost on tilemaps. Tested on map which have 30% animated water.

That thing doesnt depend on pixi, its just common sence: DONT DRAW TILEMAPS FROM CANVAS IF THERE IS AN IMAGE.

I use that optimization in gameofbombs.com , 1920x1080 tilemaps work like a charm.

wow that is some key information there nice find.