Optimizing My Minecraft Clone With GREEDY MESHING
HTML-код
- Опубликовано: 3 ноя 2024
- In this devlog of remaking minecraft in java, I add index buffer objects, vertex compression, and greedy meshing to optimize the VRAM usage of my game.
Music used:
About Town - Professor Layton And The Curious Village Soundtrack
Gant's Bar - Professor Layton And The Curious Village Soundtrack
Pokemon Black & White 2 OST Aspertia City
Useful resources:
LibGDX: libgdx.com/
Learn OpenGL: learnopengl.com/
Greedy Meshing: 0fps.net/2012/...
In the future, Please add an LOD system using a system where it samples only a few blocks inside of a chunk at large distances, with the amount of samples being dependent on the LOD level, and decreasing the further the level is. An LOD system is good for performance at large render distances, and a rough terrain generator would be amazing for generation speed with large render distances.
^
hell yeah
THIS IS WHAT WE NEED
Just use the max render distance
He said in previous video that he is gonna implement it in the future but for now there are more important things to focus on
If you do ever attempt to add rivers, I'd just like to say that rivers start out up high in the mountains as small springs.
Minecraft's rivers at first were used to divide biomes to reduce harsh borders, and they tend to be shallow. Nowadays, these are decently sized but a lot of the times these are too deep, but still maintain the same purpose. Since these still maintain the same purpose, it usually ends up circling around, never reaching to a large body of water.
Extremely psyched to see you making progress on the engine. Its really fascinating to see projects like this from an early stage.
If you would consider publishing precompiled jars of the game, I'm sure a community would quickly form, similar to the way minecraft's community did in the early days.
Don't keep us waiting too long with biomes though, I'd love to see what biomes could look like when generated with elevation in mind (thinking underground & sky biomes)
Wdym for us? That's his project.
@@ataracticDon't worry, I appreciate the excitement! If I didn't have others expecting things from me, I probably would have quit the project after the first episode, this series is about sharing the journey rather than going about it alone.
I'll work more on terrain generation after the very basics are done, i.e. survival and multiplayer, but that will take a while I'm afraid.
@@ataractic No one is uncalm? Except you
Please don't feel pressured by my excitement, it is your project after all, I'm just excited to see it happening, whichever form it might take.
I've always wanted to see a game with ore generation that follows the rules of Geology. Like if you want to find a copper vein you would look in certain rock features, if you want to find iron you would look in a bog, if you want to find coal you would look near fault lines. Stuff like that. Most games, like Space Engineers, No Man's Sky, and Minecraft, just ignore ore deposits and natural rock layers and make terrain generation simplistic and random. There are real patterns in the world, and Geology is a known science that will predict where there are certain materials. A Geologist can look at a mountain and tell you what the mineral deposits, and metals, are around the base of the mountain, under it, or next to it. He could even make some predictions about where the nearest fault line would have to be to generate a mountain that size. Or, you know, just completely random layers with no logical Geological significance whatsoever.
I agree it seems like a simple way to create fun gameplay. Minecraft in particular really did a bad effort with minerals.
One easy addition would have been diamonds spawning near lava, as the old story said. It then forces you to dig around lava.
Man I love this series ,keep going 😊
Imagine creating a superflat world and realising the whole world is just 1 face made of 2 triangles
my suggestion: LOD rendering
if something is outside of the render distance, render the whole chunk, but with each block being half the size, for a couple of chunks, then halve the size of the next few chunks, on and on until you've rendered some cubes really far away
my suggestion: using a faster language (like c or c++ instead of java; assembly would be overkill tho)
Huh? Making blocks half the size? Wouldn’t that make it a mess of blocks with gaps in-between? I would imagine that would mess with the advantages of occlusion, and also look super weird? I must be misinterpreting what you mean.
I think level of detail is mostly done by like, having separate versions of models, which have fewer polygons? I’m not sure it is applicable to something made of cubes like this where the cubes can change?
I suppose maybe one could like, somehow automatically create and store a simplified low-detail mesh for a given chunk, by maybe taking some portion of the exposed-to-outside-air corners of blocks in the chunk, and like, try to make some kind of approximation to a convex hull?
Though I feel this would be hard to make work well?
Oh! Maybe you meant, render blocks at *twice* the size (in terms of length), and render only blocks whose x,y,z coordinates within the chunk are each divisible by 2? So, (1/8) of the blocks.
I suppose that could work pretty well? Though I guess it would mean that at some distance, from some directions, the apparent distance to a 1-block-thick wall would suddenly shift by a block. Maybe that’s not a big deal though.
Oh, also, some one-block-thick coverings over a hole would disappear?
Oh, I guess that could be resolved by checking if any of the 8 positions in a 2x2x2 region, had a block in it.
I don’t have enough experience with low-level things to have a good idea as to whether or not this would work well.
@@drdca8263 You are absolutely right, the whole point of LODs is that the game already knows what goes in that location and can load lower polygon models that at a distance don't seem out of place compared to the high poly of the nearby models. In a game that has infinite terrain generation creating those LODs would require a beefy CPU and slow the game down to a crawl, the only real way to do it without tanking performance would be creating the LOD chunk after it has been initially loaded and updating on chunk modification.
@@cyclopropinon5143 C and C++ compile to assembly. Their compilers know more about optimizing assembly than most devs alive. Hand coding Assembly is a waste of time, especially with modern hardware.
@@drdca8263 Just search for Minecraft Distant Horizons or Minecraft FarPlaneTwo and you'll get what they meant (I also think they meant twice the size)
One Minecraft terrain mod I really like and makes the worlds more "realistic" is the TerraForge mod. It's something you could look into for inspiration for world gen.
*terraforged, and I have to agree, makes the terrain make much more sense and just work better
I came for minecraft optimising but stayed for the Layton soundtrack
on the subject of lakes and rivers, i've always wanted to see minecraft rivers done well, with like currents and rapids and stuff
Your voice is always so calm and the things you explain are really interesting! Please keep on doing content!
If im not wrong its a ai voice
@@stupidsniper5216It sounds like AI to me as well
@@sullivanb2306 It is, in his second video on this you can hear the AI glitch out and do an iconic loop of saying "uhhvcdsfbvbdfsvuyyhuhhhhhhhhh"
David Attenborough mixed with Delvin Mallory lol
New favorite RUclips devlogger found.
I'm absolutely loving this series -- good job! Seeing a Minecraft clone that actually looks like Minecraft is really sick. One thing I'll point out though, I'm not exactly sure how Minecraft does this, but it seems to disable culling on leaf blocks, making trees look far denser and more natural. Not sure if that'll impact performance too much, but I think it's something you should add. Anyway, good luck with this project!
Foilage has a surprising performance impact, needing to be on a separate mesh, and without self culling there's a lot more to render- I suspect I'll make it a performance setting.
@@finalforeachare you going to add the nether and the end?
@@finalforeach There must be some way around it -- maybe make the textures solid / cull them when they're a certain distance away?
@@finalforeachi am not sure but i think you can get away with ambient occlusion settings and maybe some texture duplication
OMMGGG IVE ALWAYS WANTED GRAPPLING HOOKS IN MC TY FOR ADDING!!
also, do you plan to add other custom highly requested mc features?@@finalforeach
he hasn’t uploaded in a while, that means he is working on a bigger update!!!
A quick note to people commenting: Vulkan does NOT automatically mean a "huge performance increase". If the person implementing in vulkan is not extremely well versed in how gpu's work and how to synchronize data between the cpu and the gpu efficiently, the performance will likely be much worse than even unoptimized OpenGL code. It is incredibly difficult to get these kinds of performance gains and it requires a deep understanding of the api.
Don't bother with switching; you will not get the performance as people like to think unless you squeezed every last thing out of your current implementation; and even then the gains are likely very limited unless you switch to completely different rendering mechanisms such as tesselation; which can also be done just fine on OpenGL. You'd be better off switching to OpenGL 4.5 with render objects which already comes close to the performance gains you might want without taking the massive extra step. In GL 4.5 not everything needs to be bound/unbound all the time which means there is a lot less state switching going on on the cpu side.
Gotte love the professor Layton music
You might have heard of that one semi-terrain generator thing. It starts with a grid, then picks a random tile and places it on a random place on the grid, then it has a list of what other tiles that can go next to that tile are, so on the surrounding tiles it eliminates all the tile types that won't work. It then looks at which tiles have the least possible tile types they can be, it then picks randomly from the list of tile types and that becomes the next tile, rinse and repeat. This tile generation system doesn't work as well as it possibly could though in my opinion, my theoretical solution is to take your grid which could be 100*100 in size, then create another grid which is only 10*10, it then has region types, with each tile on the 10*10 grid being a region, it then will choose what region each tile is in the same method as before, and then each region represents a 10*10 section of the 100*100 grid, it will automatically clear out some of the possible tile types each tile in these 10*10 sections can be, this will create a uniformity you wouldn't normally get with this algorithm.
This is my idea for how you could get more interesting terrain generation.
this guys minecraft clone is gonna optimized enough to run in my toaster.
I'm only a few seconds in but gosh do I love that you used Professor Layton music, such a nostalgia trip combined with Minecraft ❤
Very excited for this project, I could see this running alongside the "Golden Age of Minecraft" communities if you implement all the things you say you want to, while also sticking to the charm the old versions of Minecraft like you said in a prior video!
Good luck, love watching this!
The content of these videos and the great background music you use, I'm absolutely hooked
This is becoming my favorite devlog series on RUclips ! I'm really looking forward to the next episode ! I'm so excited at the idea that we as a community could have an impact early in the game's development ! Good luck !
this or the one about that ball. pick 1
Please continue using Professor Layton And The Curious Village music, this was my favourite game as a kid and hearing it again instantly put a smile on my face
you make this look so effortless, your wisdom is evident
Hmmm. My previous comment is missing for some reason.
In my older engine my vertex struct for world geometry was 24 bytes and it had a lot more data in it (many texture coords, a 'normal', etc.). I used bytes for x,y,z and then pushed a uniform before rendering the chunk of the chunks 'world location' to which I added the x, y and z values in the vertex shader.
My mesher was "super" greedy in that it could mesh across blocks of different types and with different ambient values.
My IBO used shorts.
Finally, I meshed each chunk into 1 VBO, 1 IBO, but six distinct regions so I always rendered only visible faces, reducing the geometry processing by nearly a factor of two.
Oh, and the last optimization I made had to do with the order in which I rendered the chunks based upon the player facing--it's too complex to detail here but cut the number of unnecessary chunks rendered down even more than using occlusion queries.
There are other optimizations in there that I'm forgetting, but I'll soon remember as I'm porting them all to WebGPU for an online game I want to make.
Looking forward to keeping an eye on your progress. Maybe I'll start a channel and do a devlog too.
I had a quick question, how were you able to make a super mesher? How is it able to switch textures across the same triangle? Does it dynamically pack textures at run time?
Also did you notice any improvements from separating your mesh into six distinct regions compared to using the in-built backface culling?
I like that your focusing on the most important things first and how you're making this because it feels like the game isn't being rushed through and the content is entertaining even if you don't know about coding. Have been looking for a devlog series like this for awhile and appreciate the work you do 🙂
This is so impressive. I am absolutely loving this series man, keep it up!
P.S.: Please add a LOD system for Higher render distances. That would be so cool!
you make this seem easy and quick despite both being untrue lol. Good luck with this project I'm a big fan of sandbox games that have great performance as I think they're the games that last the longest or at least have the potential to be long lasting, especially with mod support
Absolutely loved the Professor Layton music at the beginning, gave such a nice flashback
I understand very few of what your talking, as I don't know anything about coding languages, but I still find it all fascinating. I really wanna see where this project goes, it looks amazing so far.
Glad you didn't abandon this series so far lol.
I love what your doing, man! As a layman, it's interesting to watch you explain your thought process on the methods you use to achieve certain goals even if I don't understand the technical stuff.
I'm loving this series!
Off topic question, but I'm really curious about your journey as a programmer; it would be cool to hear about it on a next video.
Cheers!
I'm so happy to have found this project, very excited to see what is done next! I'm not a coder or anything like that, but I do know that lighting in games is really hard to implement, and it usually lags a lot if not optimized propperly.
so happy i found this channel earlier on, here for the game dev stuff and the soothing voice
I really wish your project gets more and more attention as the time goes
I think it will develop into exelent game of its own over time
Thanks for making this :)
I don't understand almost a single word what are you talking about, but I really appreciate your effort and feel really excited about this project and hope you're gonna make it!
I'm sorry about the technical talk! Future episodes should have more shiny features to look at.
@@finalforeach There is no need to sorry - it's obvious that in these kind of projects the technical stuff is unavoidable. It's even better that you are actually talking about it, because people that are into it can also express themselves and maybe give you some advice. Nevertheless it seems to become quite interesting in the future and I'm planning to stay tuned ;)
Love the professor Layton music lol
This is an absolutely incredible series
This is really cool, I can't wait to see how far you can take this project
Everybody gangsta until this guy makes mc more optimized than old mcpe
I just found your channel today and I am completely hooked, super excited to see where this project goes!
this is extremely cool as a compsci major taking java courses, your videos help me to understand the bits of theory ive pieced together from other explanations like factorio's circuits and minecraft redstone and apply them to java.
Not only a really good video, not only a really cool project, but also the professor layton ost.
Really good stuff man
I'm rooting for you, these uploads make me happy to see :)
I just started the video, and I already love the Professor Layton music in the background!
Oh my god that professor Layton bg music just sent a surge of memories back
this project looks really cool, I look forward to seeing what you're able to do with this
I really look forward to seeing how all of this cooking ends up. Love seeing what's under the hood with your explanations, keep it up!
it would be unique to play this as a character twice the scale of the minecraft player, either 3-4 meters tall or with half-meter sized blocks. could avoid constant jumping to get up slopes and change how people approach building.
I think the concept of infinite (or at least extremely high/multi-layered) terrain generation vertically is really interesting and could lead to some fun exploring! Great work!
Just found you a few days ago and I'm loving this 00 excited for lighting to make it look more like a real world
This is fantastic. Very glad I found you so early in this project!
this is already looking really good. optimization is always high on my list when it comes to Minecraft. can't wait to play it.
If adding LOD Rendering, please make it toggle able or adjustable like render distance.
Love it when details far away are still able to be seen
Microsoft is going to sue you faster than lightspeed when you release it. I love to see how well you optimise it, keep it up!
A suggestion for what you could turn your future game into: Make a Minecraft but with tech in mind! One based on the most popular tech mods, but optimized so they won't cause issues like they do in Minecraft. Mods like Mechanism (86,612,776 downloads alone at the time of me writing this) are highly in demand and even games like Factorio really lift off on Steam. I feel like we have pretty much everything Minecraft already. A Minecraft we don't have, is one for the tech lovers! Alternatively: Make a great base and add amazing modability. This way people won't be split over mod loaders and launchers etc.
All of that being said, you make great videos and I'm looking forward to being one of your first alpha customers!
Heck yeah!
Appreciate it! I've been on a Factorio binge recently, and I think the signals system in the game has more potential than Minecraft's redstone, so look forward to that! (We have a lot of features to go through before getting to that however)
@@finalforeach Personally I'm hooked on the Minecraft mod Pneumatic Craft now. It has logistics drones similar to Factorio, and programmable drones as well using puzzle/block programming. It's amazing :)
@@finalforeach, I'm glad to hear that! And good things take time, I think that I'm speaking for most people when I say that it's great to be able to watch those development update videos until we can get our hands on the first alphas. While the anticipation is definitely big (I wouldn't be a gamer if I wasn't excited for a promising new release), I much prefer you taking your time and being able to deliver a finished, rather than a rushed, product step by step. Keep up the amazing work and thanks for taking the time to reply!
love the little "catch up" videos while you work on the project with little explanations of what, when and why you do smth
even as a non dev it's fun if you are a bit educated on post processing, modding or tech in general, see you next time
Haha i love this project ! The layton reference at the beginning is so cool
This series is my fav content on the platform
this minecraft looks so good, and with that amount of optimization its going to become a very powerfull tool if it has mods
I can’t wait for this project to be finished! Even after very few updates, it looks beautiful!
i just found this and i’m obsessed. it’s giving me nostalgia from the good old days of minecraft for some reason but in such a good way. keep going bro it will blow up if you stay consistent. and i’m definitely subscribing for more.
i like your funny words, magic man
using Professor Layton music is criminal
youre gonna make me cry out of nostalgia
Parkour is my favorite part of minecraft, and feels better than most games because you dont usually slide on or bounce off of blocks when you land on them. Please keep parkouring for fun in mind while making the game! Also, a grappling hook sounds like an amazing idea!
Loving this project. I recommend also looking into Vintage Story for ideas or inspiration for your own take on the genre. Keep it up!
Are you going to implement distance-based mesh simplification so that render distance can go up? Perhaps use the noise maps to render more simply at greater and greater distances.
I did not expect to hear Professor Layton music with a Minecraft coding video, but I'm loving it!
Signal boooooost. Love your work. This is so interesting to follow along.
Adding mountain rivers and pebbly beaches would be cool, Minecraft doesn’t have these. Make beaches more shallow more commonly to.
as a student studying computer engineering and also a big minecraft fan I love this series. Dont know much about coding yet, but its still incredibly interesting seeing this project take shape and trying to analyze the code snippets you showed on screen
My day just got 10x better since ive seen you posted a video
Love the Layton background OST
this has quickly become one of my favourite series on youtube
A passionate dev wanting to take a swing at voxel sandbox game? I'm all up to follow this journey! Subbed!
Be sure to take brakes so you don't burn out. It looks amazing and I hope you will take your time without stressing about keeping up with a deadline or pressure form community. Take care!
Man, I have no idea what I'm watching but I'm loving it.. keep up the good work, and I hope you reach beyond your imagination :)
As someone who uses a samsung j2 like 26080 others, i can confirm getting an optimized version of minecraft with extreme render distance is an absolute win!
As a fan of high mobility and movement options in games I loved to hear your plans to add a grappling hook later on. Would be cool to see more things tho, both in core mechanics and character upgrades/gear items, like starbound do with movement tech or terraria with it's accessories :)
I recognize that professor layton theme anywhere! Good stuff my guy
You smart, smart man! Absolute genius. I really want to try it tho 😂
I'm with others here man, this looks cool and gives me some insight into what it takes to code a game. It's fantastic, really! Can't wait to see what the future holds!
This is such an exciting project to "keep tabs on" I'm seriously delighted by following along and it's thrilling to think of and imagine what might be coming in the future?
I'm really looking forward to in the future of this clone! It's very interesting listening to how you'd optimize your minecraft clone!
I love Starbound! I've had it for years so it's cool to hear someone mention it! I'm in the middle of actually completing it for the first time (on pause bc of college) despite having started it years ago
Have you seen the lecture of one of the Minecraft devs talking about how they redesign minecraft's terrain generation for the caves and cliffs? I found it fascinating and quite informative
I did! It was useful information, but I want to see if a different system can produce more interesting results.
Very cool video once again! It really is fun just how much i learn from watching your videos. As always, keep it up ^^
This one was a bit more technical, but future videos should have more focus on features than backend stuff.
Correction(s):
The graph at 1:33 is wrong, there are 65 numbers in each dimension total, so 0-64 on each axis, not 0-65. A classic off-by-one programmer's error.
The lack of a hard limit in height is something that could make your game very unique. Maybe you could start using different textures that give your game identity.
You could further optimize your game by removing ambient occlusion. If you use textures that shade the blocks. (Top face bright. Y facing sides side slightly less bright. X facing sides even more dark and the bottom face almost black)
That professor layton music though! ❤
I never thought I would find a fellow professor layton music fan.
I absolutely love your series, goes so in depth, thank you very much sir!
This series is amazing! And as a sidenote, you voice is really nice.
I am a teenager, and my goal is to become as cool as you one day
This has potential. I like it. The fact that the leaves are all merged like glass is a bit weird looking tho.
Yo, Professor Layton music. Haven't heard that in a bit
Loving the professor Layton music ❤
Thanks so much for reading and featuring my comment! You have all of our support and this has so much potential!
im obsessed with minecraft like games focusing on optimization, i feel like there is near infinite depth there and regular minecraft only scratches the surface!
I have 8 carefully considered suggestions and none of them are LODs.
0. Consider making the player 3x2x2 in size, more like Terraria than Minecraft, because it allows for more intricate building. The only reason not to do it in minecraft is that the digging becomes more laborious, but there are other ways around that if you violate the one-block-at-a-time principle like you mentioned.
1. A standard for streaming terrain from one process to another, so that people can write their custom terrain generators in other languages.
2. passive and hostile 3d cellular automata. Maybe make all life forms block-based except for the player.
3. if/when you add any redstone equivalent, make it more simple and intuitive, without quasi-connectivity, and with a policy of NEVER turning incomprehensible bugs into features. Do the infinite world height refactor before adding a redstone equivalent, because the behavior of redstone machines being partially unloaded from different sides is going to be a big deal to some people.
4. Cling very tightly to the simple elegance that games like minecraft lose over time, maybe by limiting yourself to a palette of 64 materials+items+mobs for all time, so that every tangible feature is carefully considered, and later in the game's life, almost all development must be shifted to intangible features like new physics/controls/commands/game modes/world generation types/new interactions with existing creatures and blocks.
5. built-in support for vehicles built out of blocks, like the many airship mods Minecraft has had over the years. But let the block grid that moves with these vehicles be first-class, so that features designed to work on the ground like pistons and crops automatically also work in a moving vehicle.
6. Occasional fear of open spaces. The huge open spaces are beautiful and freeing, and it would be great to fly over them or race over sloped surfaces, but there should be a danger. I'm thinking since many games have had blood moons before, this one could have a dangerous sun. A beautiful and peaceful night could be turned into an adrenaline rush when the sunrise is near, because some mornings, the sun rises with a shriek, and rays of light land on the player like a ton of bricks and crush them.
7. Sky islands. but at higher altitudes, some building materials can't withstand the crushing sun, so surviving at higher altitudes is tied to progressing through tiers of building materials to build shelters. Minecraft was a game about going deeper to find better loot, I want to explore the alternatives, going ever higher for better loot, and feeling lost in huge open spaces.
8. swivel grapple hook, where instead of pulling yourself towards the point you're looking at, you can freely rotate your position around it while keeping a constant distance. It would make for very interesting movement.
9. uncompressed or very simply compressed world saves that are easy to edit with other programs.
I'll be happy with whatever you do, especially if you open source it so I can do the things you didn't do. I suppose that makes my favorite suggestion the elegance one. Minecraft has proven that the game developer can never compete with the modding community on sheer volume of thingamabobs and doodads. Mods make a game great, but rarely build on each other, and never carry the game into the future. So the developer should bring stability, vision, and the most painful and necessary refactors - the things the modding community can never do.
You be the skeleton and we'll be the muscle. We will make a great team!