PIXI V2 to PIXI V3

[Hackerham]

Ok, in this test its only 2.5x in chrome: http://jsperf.com/tilemap-optimization

Can someone check it in safari?

 

[Kane Hart]

I tried on chrome for fun got some odd errors not sure what I'm looking at hehe:

http://i.imgur.com/UAVY2Fx.png

 

[Hackerham]

Ignore that error. Look at 118 vs 327, rpgmaker bitmap is 3 times slower! Other browsers work ok.

 

[Ramiro]

This will only make blts faster, I don't really know if this will improve tilemap's speed, they were already rendered on another texture.

 

[Tsukihime]

Hey, I don't mind if blt's are improved 3x.


Also, for tilemaps, I may be interested in, for example, swapping tiles...


Though that might not require any blt work since I'm changing ID's.

 

[MuteDay]

im planning on testing if you would like you can add me on skype mutation_industries@live.com

 

[Ramiro]

UPDATE:

Yesterday TiledSprites started to behave properly, after some fight with them.

Now looking why tere is flickering on sprites with custom frames... (if they use an static frame this doesn't happen)

 

[Dr.Yami]

/me is waiting for Ramiro's PIXI V3 update to work on some simple shaders.

 

[Hackerham]

I'm working on new tilemaps for pixiv3.

I'll make tests on rpgmaker, tiled and on gameofbombs. I will compare execution times for multiple ways of rendering:

0) render everything everyframe

1) render everything into several textures (current rpgmaker approach)

2) form UV's for webgl

3) form UV's for webgl using glPoints mode

As for other performance issues, game logic costs most of CPU time.

 

[Ramiro]

/me is waiting for Ramiro's PIXI V3 update to work on some simple shaders.

If you do:

this._mySprite = new Sprite(Cache.picture('img'));this._mySprite.filters.push(new PIXI.filters.AsciiFilter());It has to work... I believe, but I'm near the "stable phase" well, if you can considerer the same stage as MV "stable"... of course

 

[MuteDay]

have you fixed the water based sprites not animating right for me they are animating 1 row works ok then 2 rows wont work are stagnant then repeats

 

[MuteDay]

also would like to report massive lag when it comes to having 1) large maps then even worse with animated sprites like water

 

[Ramiro]

have you fixed the water based sprites not animating right for me they are animating 1 row works ok then 2 rows wont work are stagnant then repeats

I'm not touching the default classes's behaviur if they are broken they'll keep being broken.

At least for now, the idea is to migrate to pixi v3 not to solve bugs, that's not the scope of this repo right now.

After migration, I'll dedicate any improvement into bugfixes, but don't expect that for now...

Please, report the issues at github first, then read if they are not already there...

Keeping track of bugs on blog pots is hard.

 

[MikePjr]

I been testing this since i realized it's here, it actually works when it comes to the blending features.

I'm using it with a fog plugin, it's fine with one fog.. but if i add more than 1 layer of panning layers it starts having trouble.

I just like that the blendmodes work better.

 

[Ramiro]

That would probably be pixi's own behaviour, but yes the idea is there.

Hackerman is actually rewritting tilemap entirely, to make the animations with shaders (so they will be faster than the method used in RM)

So tilemap is going to have improvements soon.

I am just too busy with mvplugins to work with this ATM... (But I won't abandon this)

 

[MikePjr]

Ah alright. Well i will keep testing. Won't hurt to.

 

[Hackerham]

Im testing tilemaps right now, I will commit them in a hour or two

 

[Hackerham]

Ok, tilemaps are ready, please test them.

+ Tilemap is fast and doesnt use extra textures, only vertex buffers

+ Animations are handled by shaders

+ Fixed memory leak in tilemaps

+ Fixed memory leak in snap()

- Canvas fallback is slow, 5x slower than vanilla while scrolling.

* there's possible leak in TilingSprite and Sprite._tinttexture

 

[Hackerham]

Ok, shaderTilemap is a plugin now. And I use vanilla tilemap for canvas fallback.

 

[Kane Hart]

Ok, tilemaps are ready, please test them.

+ Tilemap is fast and doesnt use extra textures, only vertex buffers

+ Animations are handled by shaders

+ Fixed memory leak in tilemaps

+ Fixed memory leak in snap()

- Canvas fallback is slow, 5x slower than vanilla while scrolling.

* there's possible leak in TilingSprite and Sprite._tinttexture

Is there a way we can ever make canvas fallback 5x faster rather then slower lol? Android devices seem to be mostly using canvas.