Bevy 0.14 -- Rust Powered Game Engine 

Tbf unity is making an abstraction to ecs while bevy is *just* ecs So bevy has had alot better design decisions available.

just ECS. Yeah sure buddy.

@@dotcom4389nanite has improved a lot for foliage since 5.2

​@@dotcom4389 I heard they made it work with foliage

There are a decent number of steam games written in Rust at this point. Also, many of these are alpha/beta type projects and don't claim they should be used for large projects because they're unstable.

People who are programming in Rust are more likely to care about the systems than the games. Also, only two engines, chill.

@@techpriest4787 And you are teaching "Famous Programmer". lol.

Rust is only 9 years old for 1.0 release. That means the engines are even younger. That means it leaves even less time for the actual game. And just because it is written in Rust it does not mean that the competition Unreal, Unity and Godo does not exist. So actual game designers will stick with them. But of course your "famous progaymmor" knew all this. -.-

Bevy isn't a flagship product of a corporation, nor is it remotely as old and mature as the big boys. It doesn't even have an editor. Don't beat a toddler.

Bevy won’t “blow up” until it has hit 1.0

@@crab-cake i gueess webgpu will be the new flash games lol

@@thegoldenatlas753 i think it could be much bigger than that. peak potential is having triple a games in the browser. truly cross platform games. doom 3 has already been ported to the browser with wasm and webgpu will make it possible for new games.

Bevy can build to WebGL and run in browsers with WebGPU support

Ive done alot of talking in the editor dev channel and things have finally kicked into gear it seems

Yeah, once bevy gets a good enough editor I might go and do a project with it. For now I'll stick with c++ and unreal

Honestly I was thinking that was more imoressive than the game engine itsrlf

about the same kind of thing you'll experience using any non game engine framework

I mean now things are pretty well developed. Curious what things you couldn't do now.

1.5 years ago is a LOOONG time ago. I tried it this month and the ecosystem met all my needs.

@@askeladden450 true. Rust ecosystem is evolving at a really rapid rate. Even an experience from 6 months ago isn't relevant anymore.

That's the case with any non-mainstream games language, gamedev it's very tied to C++

really? pretty sure you could achieve most of that stuff in godot. other than the motion blur. idk how to do that one

@@RenderingUseri think it can do most of them tho they dont have per object motinblur still to my knowledge so u can’t have montion blur in godot that doesn’t suck

Yeah, I agree. It's such a letdown when you consider how many years Godot has been in development...

They're focusing on advanced flashy features to get users instead of adding basic functionality, such as an editor.

@@autismspirit OK, but that still doesn't explain how can Godot still have such poor 3d graphics in general.

@@autismspirit an editor is not really *that* basic. You need functionality before having UIs to use them.

@@autismspirit This was made as an api for programmers to use. Being forced to use an editor doesn’t give you full control. There are many game frameworks like this.

@@beetledjuice3062 A rudimentary GUI is way easier than virtualized geometry, for example.

FXAA is to me a better implementation of MLAA than original SMAA, i would argue; if it's too blurry just tune fxaaQualitySubpix down and other parameters. And TAA has nothing to do with FXAA and MLAA at all. Intel's CMAA2 is quite nice btw.

@@SianaGearz Is TAA not a direct successor/offshoot of FXAA, or am I misremembering?

@@HickoryDickory86 TAA is accumulation antialiasing, where the TTA controller injects a differing subpixel offset into rendering of each frame to reveal sub-pixel details over the course of several frames and then averages these several frames to provide true anti-aliasing, while trying to keep ghosting to a minimum by forward-correcting prior frames using motion vectors. Basically distorting the old picture to align with the new one, and then averaging them out. It doesn't have to turn out blurry, but due to issues in frame to frame coherence of the underlying renderer and numeric precision issues, usually ends up being blurry. Say in practice, it doesn't actually add a limited number of frames together, but maintains a rolling decay buffer, which accumulates long term error which imprints as blur, so any imprecision in forward projecting prior frames lingers for ages. Fundamentally the forward projection has a large error, since it's effectively snapped to the pixel grid, does not maintain subpixel accuracy. This is also what FSR2 does better, since it can do the forward projection at a higher resolution, but there's also a lot more other trickery, one could write a book about these things. FXAA is a form of morphological anti-aliasing which works on a single image with no further information. MLAA starts off the observation that if you have a fully aliased rendering of a black half-space adjoining a white half-space, and compare it with a fully anti-aliased one, you find that you can accurately reconstruct the grey-tones just from the stair stepping pattern of the aliased picture. Namely you can do this by a directional blur filter and apply such a blur filter vaguely perpendicular the edge direction, with filter parameters that depend on slope computed from the stair step pitch. The algorithm is further augmented with detection of features that need to be processed, to make it work on something that isn't a pure monochrome picture. As such it ends up blurring all high-frequency detail in a real picture, because the picture really doesn't look like the algorithm-underlying observation. Since morphological anti-aliasing does not have access to any sub-pixel information, it is not a true anti-aliasing algorithm. I'm very sorry, describing image processing algorithms in text without pictures is really really difficult. But in case of morphological algorithms, Intel Developer website on CMAA and CMAA2 should be helpful. Some morphological algorithms take into account say depth buffer or other intermediate image buffers to reject a bunch of pixels that would otherwise be blurred and keep them sharp. But this runs into the issue that many shaders produce internal aliasing that gets missed. Also you know what, i am now seeing that SMAA was much nicer than i remembered. But in the form where it can shine, it was also a substantial performance hog for the time. FXAA solved the performance problem and then it just stuck as the default implementation, because you can mindlessly throw it on things. All of these MLAA variants have their own detail issues.

CMAA2 is open source under Apache 2.0 license? Or maybe you meant specifically CMAA 1?

@@kai-.-man I had no idea! Thanks for telling me! Now I wish Bevy would roll CMAA 2 into their engine.

that _is_ on the todo list before version 1.0 release ill start properly using it in production then.

For now there is a tool to make blender into a game editor for bevy

@@thegoldenatlas753 so thats what aarthificial is using

@@RenderingUser i think so, there's several options available rn

This was made as an api for programmers to use. Being forced to use an editor doesn’t give you full control. There are many game frameworks like this.

It is for optimizing high-resolution models for performance by dynamically down-scaling them, as needed. It is very inefficient to render triangles that draw just a couple of pixels in the screen, or any at all.

​@@Rodrigo-xf2oe Thanks. I'm trying to deepen my understanding of these concepts (and my understanding of moden graphics pipelines is very elementary). My current understanding is that the key design principle behind meshlets is to more efficiently handle large detailed meshes. By being able to address sub meshes of a much larger mesh you have opportunities to more efficiently work with them in a pipeline, including but not limited to improvements in how you do level of detail. Other efficiencies gained with using meshlets could also include the way the geometry information is batched, culled, streamed / moved to and from host and GPU memory etc. So my primary comment wasn't that LOD or 'down-scaling' isn't important, just that it wasn't the only (or even primary) motivation for introducing them into Bevy.

@@jamad-y7m but it isn't rust enough for rust programmers

in my experience doign exactly this with Bevy it works well. the ECS abstracts a lot kf borrow checker woes away from you, which is very useful. Other than that, you can easily just spawn a bunch of entities and go ham with 1 big mega function even though that's not intended/ideal. Nannou, a creative coding framework in Rust is also switching to being based on bevy.

@@moonskined Tiny Glade but they have their own renderer they use, they have upstreamed some of their work back to bevy tho.

the most high profile one is called tiny glade. Though as far as I know they do use a lot of custom solutions on top of bevy.

@@lolligerjoj1 Tbf that's a selling point of bevy, its modularity and ability to be customized is big focus.

@@thegoldenatlas753 I personally agree, but probably not everyone is looking for that, I think a lot of people prefer a "batteries included" engine. To me, it's the best engine right now.

@@lolligerjoj1 Yea? I didn't say bevy *wouldn't* have stuff, like rn there's talks to upstream a popular UI crate that was made for bevy. Alot of this 0.14 update render changes are upstreamed from tiny glade. Etc etc.

??

As far as I get from googling there is occlusion culling in bevy, but it's bugged 😕

0.14 is the biggest release yet though, and things are expected to iincrease with 0.15

its fine you have external editors sometimes that can be even better take LDtk for example

This was made as an api for programmers to use. Being forced to use an editor doesn’t give you full control. There are many game frameworks like this.

Rust's type system is adequate to give you the expressiveness of an editor with the level of control you get from code

From devs that work on editor( a comment on bevy discord) :" A basic (but official) running prototype within the next four to five months seems likely. But I wouldn't put money on it. " Basically it's hard to say when exactly ...

Hold on, now the rusty bevy missioners will show how you don't ever need an editor. And just looking into sacred rust code should be enough for you.

@@MikAlexander its made for Rust because Rust is better than C.

@@thegoldenatlas753 Rust is rusty, C is citrus fresh ;P

​@@thegoldenatlas753Rust definitely isn't better than C

@@thegoldenatlas753 absolutely not lol, not for games at least

@@aamorous why

Say that to Three JS, there are many game engines made to be a programming api for programmers to use.

@@samuelhulme8347 I'm talking about UI to use in the game. For example, libGDX has it. Bevy only has the bare minimum, making anything more than a bunch of buttons with it is quite tedious

@@SomeRedTeapot new UI stuff is being upstreamed from a crate that's popular with users, editor work is waiting on the bevy remote protocol and reflection In meantime people are building editor prototypes to flesh out current needs

@@thegoldenatlas753 That's good to know, thanks!

Following the development more closely (as in actually reading github discussions etc.) its very clear that a lot is being done towards better UI. Bevy always does a new release every 4 months with whatever is ready and it just happened to align so that UI things will be in next release most likely

is this so? i wouldn't be so sure. From what i can tell a organized project like bevy is capable of catching up, by which i mean implementing things faster than the older/bigger engines can add more features. Onba high level, this is mostly due to lessons learned and a fresh implementation being cleaner due to it. And bevy is not just 1 persons side project anymore, theres 2 fulltime jobs, many Subject Matter Expert roles, and many more people contributing the features they need.

@@laundmo bevy is awesome. But again it's dwarfed by big boy game engines. Even Godot (the arguably most promising underdog) took 10 years to get to where it is.

Why?

​@@thegoldenatlas753because of the compilation times? The case is even worse than C/C++ build times.

​@@thegoldenatlas753If you have a small single threaded game, Rust is going to be cpu bottleneck very fast because of amount of cpu instruction rustc generates, C is basically macroses for assembly, lightweight, no compiler guards and fast

@@aamorous rust compiles down to assembly just as well as C They are constantly being compared with little speed difference between them

@@thegoldenatlas753 you don't get the point, I'm not talking about speed. Rust is known for making big binaries, there are more instructions compared to gcc, it's because of 'bloated' optimizations compiler inserts in final executable, it makes cpu bottleneck. Speed always comes at a cost.

@@ar_xiv you've never actually compiled a rust project have you?

i get, with some setup, sub-second iterative compile times. i'm perfectly happy with 0.8 seconds.

This was made as an api for programmers to use. Being forced to use an editor doesn’t give you full control. There are many game frameworks like this.

@@samuelhulme8347 This is untrue, because bevy does want to be a full fledged game engine. It is not on it's 1.0 yet so chill out, and their plan is for the editor to be optional.

bruv the ecosystem is fairly new give it a second

Tiny Glade just released demo recently.

That's a non-argument. Rust is still young and as in any other industry there's massive inertia when it comes to commonly used programming languages. Being able to choose a language purely on its merits is a rare luxury.

In what, < 10 years of popular and mature existence? Surely you are being fair.

This was made as an api for programmers to use. Being forced to use an editor doesn’t give you full control. There are many game frameworks like this.

@@samuelhulme8347 i mean, they are working on an editor now so that's not entirely true. they said it should be ready in 4-5 months.

@@samuelhulme8347 sure, I wouldn't call it game engine though

@@FengWan-v8y why would you not call it a game engine? An editor is just an interface that interacts with the game engine. Scripting is just another type of “interface” that can interact with the engine. Look at Minecraft, they use a custom engine (based on LWJGL) and it doesn’t have an editor.

@@samuelhulme8347 I'd call it library or a framework

No, you can use SDL or SFML in Rust just fine. Bevy is a full engine (like libGDX) with a focus on ECS for the architecture.

Honestly RU-vid stats have always been off, but lately they've been just useless. Often for the first 5 or so minutes I often get 0 analytics data at all... no clicks (but I do gets likes and comments). Makes it very hard to make early judgement calls about thumbnails, titles, etc. But yeah, don't put any stock in the views of a video for the first 24 hours, it's always off. I don't think I would compare it to SDL or SFML, it's a layer of abstraction higher than that. NotGate's LibGDX is a pretty good analogy. ORX is another, as is The Forge.

@@notgate2624 Oh I see. Thanks for taking time to correct my conjecture. I'm very new to computer science and game stuff so I don't know what libGDX is but I'll read up on it. My idea of a game engine, at least so far, is a kind of super same where you can swap out assets/mod it to produce a new game. At a glance, to me, Bevy doesn't quite fit the bill but then again, what I'd classify as "Monolith" engines like unity have pretty much messed up my perception. More research for me, lol. It's a goal of mine to produce one of those some day no matter how useless.

@@gamefromscratch Well, I understand that much. The x views in y minutes is a recurring joke in the youtube community 😅. I don't think most people believe it since me of all people understands the idea. So when you say it's a layer of abstraction higher than that, does that pretty much mean that it might use something like sfml/sdl in the code except it's more organized for the purposes of aspiring game devs? or something along those lines??

Chill motion blur can be good if it per object instead of camera, on nanite well a poop igpu could handle billions of polygons with it at 20fps, so it not so bad .

nah, it's pretty good

Said BrentMalice

@@diadetediotedio6918 yep thats wat i said. congrats on reading

why

Why so malicious?

@@NapkinKing FOSS would be better if there wasn't 30000000 "options" for a program, and instead there was atleast 1 that didn't suck. everyone wants to feel super special and make up their own thing, and then demand donations and get salty when no one pays money for something that only has half the features anyone could need. same thing with kdenlive. anyone who worked on kdenlive is a moron who wasted thousands of hours for literally nothing. their life was wasted for 0 benefit to anyone.