BitOfGold

Members
  • Content count

    259
  • Joined

  • Last visited

  • Days Won

    2

BitOfGold last won the day on November 7 2016

BitOfGold had the most liked content!

1 Follower

About BitOfGold

  • Rank
    Advanced Member
  • Birthday 11/27/1980

Contact Methods

  • Website URL
    http://www.bitofgold.com
  • Twitter
    BitOfGold

Profile Information

  • Gender
    Male
  • Location
    Budapest

Recent Profile Visitors

2,246 profile views
  1. Energy.js playground

    @PhilT Thank you, this is cool! Good to see a base comparison between engines! I was searching for this for weeks. Now we have several comparisons, yours, @lo.th 's Labs... https://lo-th.github.io/Energy.lab/#basic http://lo-th.github.io/Ammo.lab/#basic Ammo's broadphase is very good when less objects are moving and more are static (jumps up to 60FPS when bricks stop moving) It is ideal for games, as this the frequent case with a few moving characters, objects, triggers and many static elements. A very good signal is that playcanvas uses Ammo now.
  2. About performance optimization

    1. because the engine loop calls rendering on requestAnimationFrame, so it's synced to a 60Hz monitor. So its a good thing, the rendering is faster than needed. Try it on a slower machine or add some few hundred meshes and physics, you quickly run out of that idle time 2. This looks really fast to me, the graph is 0.7ms from left to right... at 60Hz you have 16.777ms for every frame, so it's nice. You will only now what to optimize when you got a really high load. I try to test everything on my 3yrs old PC with GTX970 and i7 4790K, and my 6yrs old PC with Radeon 7970 and Phenom II X6 1090T. So older and older machines, new and old phones etc.
  3. @AJTJ My insight is that this is a pretty hard theme, just dont give up Multiplayer can be the most challenging thing but it's really cool to see when something is starting to work! @Raggar This is a really cool example, seeing client side prediction (with stored inputs). I'm making a physics based multiplayer game too. It's a bit early so nothing to show yet. It is based on Colyseus, but a bit modified version of it - I have many objects in the world and one of the challenges is the spatial hashing of objects so i can find and only send the closest ones to the player. The physics engine is the WASM version of Ammo (Bullet port). This runs really fast, nearly native speed, can handle a few thousand (!!!) moving capsules and about 1000 static objects on a big world. at 60 ticks/second. With single player, I use a Worker with the same game logic that goes into multiplayer. The Worker uses WASM/Ammo too (in the browser). This is much slower (4x-5x), but can handle and synchronize about 500 capsules at 60FPS. The bottleneck is with javascript workers the data transfer speed, even if positions are spatially filtered and sent only at 20/sec. (like in multiplayer) It's a shame SharedArrayBuffer is disabled because of Spectre and Meltdown!!!!! Also, the worker version is much slower in Edge (I think the data transfer is slow, but the physics engine is slower too), Chrome and FF is faster.
  4. Energy.js playground

    @MackeyK24 I'm using Ammo through colyseus.js on the server side, and in a separate worker script with game logic in browser. It's really different from a BABYLON plugin so I think i cannot make a plugin from that. (It's mixed with game logic) Ammo is really fast with the WASM version of it, I really don't know Energy js is faster or slower. Most of my problems with Ammo/wasm is its really easy to make memory leaks and then it chrashes. Yes it's tedious to interface with Ammo/Bullet, but its several times faster than cannon.
  5. Energy.js playground

    Just when I finished integrating Ammo.js into my project ... (crying inside)
  6. How to achieve a better SSAO effect?

    I don't know why but i like this scene... https://playground.babylonjs.com/#FUMYH5#6
  7. How to achieve a better SSAO effect?

    With SSAO 2 and defaultpipeline 's fxaa antialiasing it looks slightly better: https://playground.babylonjs.com/#FUMYH5#3 But, SSAO is one of the effects (like chromatic aberration or Depth of field) in my opinion, that you just can't 'enable' and you are OK, it has to have an artistic value, or a good setting that means something in that scene, like with a low poly setting it emulates a bit of global illumination (darker in cavities) or enhances cavity or details of vegetation. Most of the time it just makes things worse than before. (For me ) ssaoRatio is the size of the effect, so 0.5 means half resolution.
  8. DirectX Raytracing in babylon

    Yeah you can buy this monster, its 25% off now (And Its G G O O L L D D ) https://www.nvidia.com/en-us/data-center/dgx-systems/
  9. @dsman That Three js bloom is really cool...
  10. heightmap and physics problem

    Yes, even cannon heightmap is too slow (for my game, 8 vehicles + many obstacles, powerups and bullets, on a 4-5 yo laptop it runs with 4fps, on my PC 60fps) I'm now working on Bullet physics on the server side (Node-js , with the wonderful colyseus library) My code is here but I can't get it to work. (It works better, fillse the holes and bumps, but I did not use the physics plugin, so this needs some rework) Maybe after some sleep... https://www.babylonjs-playground.com/#131CZC#37
  11. heightmap and physics problem

    This loads my little racetrack. SPACE starts the heightfield debug (Take caution it can freeze the browser for a minute) https://www.babylonjs-playground.com/#131CZC#34 As you see from the image, the cannon heightfield is not usable, it has bumps and objects fall over.
  12. heightmap and physics problem

    Here it is, press SPACE to show debug mesh on ground. https://www.babylonjs-playground.com/#131CZC#27 This shows that create cannon heightmap works well with a groundmesh (as expected)
  13. heightmap and physics problem

    @RaananW I agree, Cannon heightmap is good - the problem is with the algorithm that makes a Cannon heightmap from a mesh. I made a racetrack with blender, sculpted and about 40k vertices. The vertices are not at all at a regular grid, and the mesh is not rectangular. Maybe that is the problem, BABYLON's ground works much better in his situation! The resulting Cannon heightmap was not at all usable, because it had "bumps" where the plugin filled the (sparse) heights. I'm trying to make an example, because I ported a code from cannon to SHOW the heightmap: https://github.com/schteppe/cannon.js/blob/569730f94a1d9da47967a24fad0323ef7d5b4119/build/cannon.demo.js#L977 Also I made a better create cannon heightmap that fills the "holes" in the heightmap with a variation of Dijkstra map. It works much better (for me), maybe someone can use it for a similar problem. Showing is the best telling so I'm back in a few hours...
  14. heightmap and physics problem

    About the heightmap: the BABYLON physics plugins make a heightmap body out of the "ground" body, but only an approximation, so it only works if the terrain is "soft" (no big height changes). I think it is better to use boxes for the maze, and heightmap (or a box) for a "soft" terrain.
  15. heightmap and physics problem

    @Wingnut thanks for the mention, this is the maze playgorund: http://www.babylonjs-playground.com/#LYCSQ#250 I started it but its now at version 265 I think? Others are re-working it.