/gedg/ Wiki: https://igwiki.lyci.de/wiki//gedg/_-_Game_and_Engine_Dev_General
IRC: irc.rizon.net #/g/gedg
Progress Day: https://rentry.org/gedg-jams
/gedg/ Compendium: https://rentry.org/gedg
/agdg/: >>>/vg/agdg
Graphics Debugger: https://renderdoc.org/
Best AI game dev: https://grok.com
10 games made with grok 4 heavy (only 300$/month)

https://x.com/EHuanglu/status/1944708868766105863


Requesting Help
-Problem Description: Clearly explain your issue, providing context and relevant background information.
-Relevant Code or Content: If applicable, include relevant code, configuration, or content related to your question. Use code tags.


Previous: >>105911078

>>105943307 (OP)
Is grok the best for learning OpenGL?

>>105943325
grok's the state of the art for intelligence and research. their coding model releases in august for super heavy subs 300$/month. at the moment claude opus is still better for coding.

Groks waifu coded my next hit game!

fuck off elon

>>105943412
take your elon derangement syndrome and fuck off to /pol

How do I destroy a DirectX9 context?

>>105943793
Create an OpenGL context and let them fight it out

>>105943793
easy questions like this your better off asking grok 4, 40$/month. he'll give you better answers than sarcastic tards on /g/ >>105943804

>>105943845
wow look who cant take a joke

>>105943854
this is not the place for that.

>>105943854
this is super serious thread, please no make joke

>>105943854
We are conducting matters of the utmost importance, you fucking fahgot.

>>105943936
Go pay more people to pretend to play games for you Elon

>>105943949
Elon is sleeping right now.

>>105943949
kek, elon derangement syndrome is strong with this one.

>no source code
am I just to believe it can do this things

>>105936010
you wish. Every compiler relies on vtables for more flexibility than the slop you suggested when it comes to having methods in structs.
>>105936052
>>105936065
lmao
So you are gonna dodge your own takes about vtables and focus on generics now?
Keep ignoring your own takes and get focked in your shit lmao
>>105944104
>me looks inside
>finds a game's sauce
https://github.com/RodPaDev/space-x-slingshotter-game

>>105943325
>Is grok the best for learning OpenGL?

Just use learnopengl.com

does anyone have the short version of how to use unreal engine without epic launcher? build from source on github?
i'm pivoting from touchdesigner to unreal engine for realtime 3d visuals (ie bands triggering 3d graphics in realtime, no pre-rendered video)
i also am curious about game design but am debating either Love2d (Lua is the only language I really kinda partially know) or Godot (Road to Vostok dev pivoted from unity to godot and it's quite impressive), but considering i am pretty dedicated to UE for visuals should i just go down that route? apparently devs have made entire games from blueprints alone so with my limited programming knowledge maybe that's also an option?

>>105944319
eat shit and learn Vulkan, openGL is deprecated on apple shit entirely so it's either that or Metal

>>105944572
no thanks i will wait for directx 13. hope it wont suck this time.

>>105944644
open standard vs proprietary, enjoy sucking microsoft's cock, but you're not trying to ship on multiple platforms probably

>>105944572

I want to pay crypto to people to stream my game. What is the best way to do this? Will I get banned for this from 4chan, reddit, twitter, etc? What about blogger? I don't want to make a website just for this.

My idea is to have some place where people can post their VODs and I post proof of the transaction.

>>105943325
Of course not, come on.

>>105945253
Just pay them normally you stupid fag.

>>105944564
>how to use unreal engine without epic launcher?
unless you are planning to ship a game, its not worth it. I stopped using unreal every since 5 came out because my GPU got obsoleted for the engine. UE 4.26 used to take 100 GB to build from source. I used to disable unnecessary shit live VR but the runtime performance wasn't much different from the release builds you get from launcher so I'd say don't bother with it.
>i'm pivoting from touchdesigner to unreal engine for realtime 3d visuals
unreal engine is only good for visuals. Using both is also an option, generate audio in touchdesigner and visualize in unreal, best of both worlds. You can do that with OSC blueprint plugin that literally takes care of everything for you. Just setup an OSC port in touch designer or any DAW(almost all DAWs support OSC) and you should be good to go. Here is one video where a guy creates tracks in Ableton and use Max to send those track data to unreal via OSC and visualized that data unreal engine in real time.
https://www.youtube.com/watch?v=0QdOqG0BBAI&t=240s
>should i just go down that route?
my personal openion is yes. Niagara does some sick shit with ease that will be very hard to achieve with godot or even unity. You can create very cool visuals for your music with niagara. Not 2d but actual breathtaking 3D visuals where objects, lights and textures react to your 808s
https://www.youtube.com/@shimlaDnB/videos

>>105945253
im sorry but are you retarded or something?

>>105943307 (OP)
OMG BOYS, IT CAN THROW A TEXTURE ON A CUBE, it's so OVER

>>105943355
>>105943325
>300/mo
so im priced out, good to know

> This was entirely created by Grok 4. This is my benchmark to test models and I never expected Grok 4 to be able to do this, no matter what I added next it just accomplished. Sometimes I needed to try again or change the prompt ever so slightly but here it is.
It would shit itself in a second if i tried anything remotely fucking specific, everyone knows that for a fact.
Shills have 0 shame.

>>105946242
try it for yourself, ai doubters are going to get destroyed by ai adoptoors in the very near future. and grok 4 heavy is the best by benchmarks and reviews thus far

>>105944312
>Every compiler relies on vtables
Runtimes use vtables, not compilers
Vtables are only needed for dynamic dispatch, if no inheritance is used then the method call is static and the method can be located by the compiler, no vtable neccessary

>>105946304
i don't think i will thank you very much

>>105946304
>ai doubters are going to get destroyed by ai adoptoors

More like everyone is going to equality be destroyed all at once, just one group was more oblivious to it.

AI still hasn't been able to write non-trivial code

>>105946304
>in the very near future
just 2 more years!

>>105946476
OpenAI model took second place in a programming competition yesterday, this is literally the last time a human could win something

>>105946538
>programming competition
I said non-trivial code

>>105946538
so grok is just a lamer deep blue. Cool.

>>105946538
>programming competition
not real programming

>>105946680
half of code generated at google is ai chud. your ignorance is not an excuse for stupidity.

>>105946877
>google
even more fake programming

>>105943355
>at the moment claude opus is still better for coding.
LLM shits out useful code maybe 5% of the time and it still requires lots of manual changes. If you use LLMslop in any other capacity, you're going to be replaced by it before long (and no, that doesn't prove its good, it just proves shit quality software suffices in many cases).

>>105946877
>Company betting big on AI makes unproven PR statement about AI
Fascinating.

>>105947237
AI is now the CEO of Google.

>>105946209
my cube is still untextured, naked, and vulnerable

>>105945420
I don't have normal money

>>105944312
>https://github.com/RodPaDev/space-x-slingshotter-game
The game loop is extremely simple, other LLMs could have done this already. From what I see, I say good luck maintaining anything large without comprehension yourself.

For 2D shit what's the best library to use? Is it worth using SDL3 GPU for 2D?

>>105946327
>compiler doesn't care about a features that is used at runtime
you do realize that this is a completely retarded (You) problem, right?
The frontend AST that is needed to do implicit changes you are so crying about is also a part of the compiler itself.
Focusing on irrelevant compiler vs runtime usage of v-tables even when no one said compiler "uses" vtables is... as expected of a typical troll, I guess.

>>105947760
SDL is the ideal schizo / masochist / superiority complex option, you will fit into this general extremely well!
godot is ideal if you want a GUI (but C++ does have imgui, and drawing your own UI is soulful). Uses gdscript mainly, technically Godot has C++ support, but you are going to be reading tutorials in gdscript, and chances are gdscript will run at 200fps because most successful games don't do heavy lifting logic (the reason for FPS drops is more likely due to godot not being very good for 3D / non optimized meshes).
unity is ideal if you want to pivot into 3d/2.5d, and C# (or whatever unity uses) is probably better than gdscript, and if your game became popular C# can be decompiled by the community and modded (that's how most indie games support mods).

>>105947660
then pay in Pizza?
1 pizza for streaming at least once.
1 complimentary pizza for finishing the game.
>>105947723
>I say good luck maintaining anything large
that game itself is large for web devs to maintain.
Imaging have a retarded, heavily financed virtual machine at your disposal that uses two language, one for markup and the other for layouting and coloring along with a third language that handles UX. But all that is so shit that web devs have always been on constant race against one another for creating "the" new framework where all of them do the same exact thing but get marketed that this time, things will be different.
Web shits are truly insane.
Meanwhile we system devs dabble with our in-house renderer and I/O manager like its the norm but get paid peanuts compared to web devs. Truly bizarre.
>>105947760
>what's the best
raylib
>Is it worth using SDL3 GPU for 2D
we aren't there yet
2 more weeks

>>105947760
people here seem to like raylib.

>>105947760
headless Godot

>>105947760
SFML is easy to just get stuff drawing for 2d

>>105947770
You said that method calls require vtables
Vtables are a runtime abstraction used for virtual dispatch
If there is no virtual dispatch, there's no vtable, and the method is resolved at compile time

Is this a good site for learning to code?
https://www.learncpp.com/

Can't be bothered to reply individually, so as the anon asking about SDL3 GPU...
I'm not actually interested in making games, per se. Engines maybe, and other perhaps game adjacent stuff, like just interesting graphical things like a virtual aquarium and then other content creation I guess programs — digital painting,3D modeling, and so on.
Interested in accelerated 2D graphics for the painting/drawing kinda thing, and obviously eventually 3D too but I figured I'd start "simple" but I also want to learn something that will be useful beyond that. So that's why I'm not interested in godot or an engine.
I've looked at Raylib and it seems simple enough, but (and this is *very* important) will it be good enough for the imaginary future where I actually build professional level content creation programs? Don't want to prematurely handicap myself lol. Probably wouldn't hurt to make some simple games along the way though and Raylib may well be the way to go for that. I've also used SDL2 and SFML a bit in the past with C++ and sometimes C (for SDL). Figured the new SDL3 might have greater potential though.
Also for GUI I guess I'd go for imgui or some other similar stuff. There really don't seem to be any good retained GUI options unless I made my own (why not, I could interface directly to the X11 protocol right? lmao).

Anyway I'm using Odin if that makes a difference. Figured I'd learn a "modern" language.

Anyway, I'm sure I'll have real progress to show from my endeavors real soon now™

Oh btw are there any good explanations of the graphics pipeline (newer than that one from 2011) or other shit that explains the general idea of how all this shit works? Like a high level overview.

>>105948129
For learning to code in general, I can't say. For learning C++? Yes. Better than any of the books and paid resources 2bh.

>>105948321
learnopengl.com is your best resource

>>105948359
But OpenGL is deprecated and I like the Vulkan tutorial better, but neither is what I would describe as a "high level overview".

>>105948321
I guarantee you that pretending like anyone cares about the tools you make is a trap that always ends poorly.
I know that every engine dev says they don't do it for the money ("it's going to be free when I release it!", maybe with a little bit of "it's under GPL but if you give me a royalty I'll let you sell it~") but in reality, unity and godot IS engine dev with benefits.
The difference is the end goal, its either you make a game or you don't. You could make a game with an engine using all the conveniences of the built in level editor and default GUI and it will feel like a "Engine" / "Shovelware" game.
Or you could write your own map editor or physics or use Odin/C++ or GUI using primitives / low level API or you could make a "Engine Plugin" so that you could give / sell your code for other people to use.
Legitimately, you would be much better of learning shading tricks on an engine so that you could make your own Virtual Aquarium than use Opengl/whatever, because writing your own GLSL shaders is absolute torture, and you might actually prefer raytracing because you like having volumetric rays and other shit.
>No I won't!
But you can add volumetric rays without raytracing using the same engine, there are probably like 1000 tutorials about it on youtube, maybe 1-2 for SDL+opengl, and absolutely 0 for odin (have fun translating C++/OpenGL to Odin/OpenGL).

>>105948454
"Newer = better" is not the approach you should be taking into programming, and that's as high a level of an overview as you're going to get

>>105948483
Vulkan is better, though.

>>105948507
From your perspective no, you're just putting in more effort to get the same shit done

>>105948515
Nope, nta also but Vulkan is better. Just because you can't into verbosity doesn't make it bad.

>nvidia 4050 doesnt support parallel queue execution
excuse me wtf?

>>105948525
Vulkan makes you do more work for no benefit

>>105948621
>for no benefit
Yeah if you're making 2D shit. I understand, though you have an extremely narrow mindset.

>>105948637
Nothing you're making in 2D or 3D will see tanigble benefit from using Vulkan

>>105948454
I can't wait for you to learn Vulkan just to figure out Vulkan is just Opengl but with a bunch of extra steps with zero conclusive FPS benefit (except for intel GPU's I think? but I think those old games don't use GLSL, they use ARB Assembly Language I think?).
The semaphore thingy (multi threading your rendering, this isn't unique innovation for vulkan, it's basically following the same rules opengl 4+ follows, but C++ doesn't have the semaphore thing built in) / mesh shaders (AKA, sort of like compute shaders for generating meshes? +Nvidia has mesh shaders for opengl), and maybe some shader compilation speed bonus (emulators with vulkan tend to compile shaders faster than with opengl, pretty much the same shader language by the way) but that's basically it, none of those features give a huge FPS benefit, like I saw GDC talk about Alan Wake 2, and the mesh shader were used for runtime occlusion culling I think, but it only saved like less than 1% of FPS (makes sense because of the worst case outdoor scenario, it's funny how when you have an outdoor overworld, there is no magic optimization that matters other than frustrum culling and Level Of Detail optimization, because 90% of the objects are being drawn, you can't hide the objects).

>>105948651
>higher performance is not a tangible benefit
okay, retard

>>105948677
higher theoretical performance, you aren't making AAA games
so sick of this conversation every single gedg thread

>>105948681
Mad cause bad. Not my fault you're retarded.

>>105948690
You haven't made anything, shut the fuck up

>>105948694
Stop projecting your failures onto me. No need to be jealous.

>>105948469
Fair enough for the virtual aquarium, for that an engine would be the most direct route. But the proiblem is I don't really care so much about the end product. Like it would be awesome if I actually made something, but the projects are really more of just an excuse to learn shit, since learning shit and (at least attempting to) make shit is what I enjoy. Also, though obviously I'd want my fish to look good (why do shaders have to be so gay and complicated? Why can't they be easy like shaders for 3D renderers so people in the community can have good jobs?) I'm more interested in using it as a test bed for esoteric fish AI and other shit like cellular automata or whatever other stuff I find interesting but need some reason to actually learn.

>But you can add volumetric rays without raytracing using the same engine, there are probably like 1000 tutorials about it on youtube, maybe 1-2 for SDL+opengl, and absolutely 0 for odin (have fun translating C++/OpenGL to Odin/OpenGL).
lol yeah, kinda stupid to learn a new language and other stuff at the same time when there are few resources. Anything in C is simple enough to translate though and it comes with bindings to all the popular stuff (sdl2/3, raylib, sokol, wgpu, etc. which are all rather readable). C++ of course would be more complicated if they bust out OO stuff.

Anyway yeah I know I'm an idiot and will likely never accomplish anything. It is what it is.

>>105948698
hows the registry dumper coming along?

>>105948666
satan trips

Anyway I've gone through the vulkan tutorial, it's tedious as hell but that in some ways makes it easier to see what's going on. I certainly don't know vulkan... Everything made good sense as I did it, but there's so much different shit it's hard to understand the bigger picture (without more practice and familiarity). But all the details is kind of what I wanted. I don't want shit to be "magical" or hidden behind opaque procedures, because then I will never actually understand.

Of course I also don't exactly want to deal with the verbosity of Vulkan, thus SDL GPU and even though it hides a bit I at least, having gone through the vulkan tutorial, have some idea of what its doing for me.

>>105948730
>I don't want shit to be "magical" or hidden behind opaque procedures
That's exactly what Vulkan is doing for you

>>105948677
I literally guarantee that your naive vulkan implmentation will be slower than a naive opengl implmentation. vulkan has a thousand footguns.

>higher performance
remember, this is only about CPU driver overhead. if you are GPU limited (most likely), 1ms more of CPU overhead doesnt matter. and its piss easy to reduce overhead in opengl to the point where it doesnt matter. use a few big buffers, upload at the start of the frame. render with multidrawindirect. done. "opengl is le slow" is from retards that spam 1000s of draw calls instead of just sorting+batching.

>>105948755
>I literally guarantee that your naive vulkan implmentation will be slower than a naive opengl implmentation. vulkan has a thousand footguns.
Shut up rust tranny. My feet may be blown off but at least my dick is intact.

>>105948755
I can issue OpenGL 100,000 drawcalls in less than 0.1 milliseconds

>>105948769
i doubt that. are those indirect/mdi calls? doing 1000 simple calls with no program/texture switches is like 0.5ms for me.

>>105948848
>doing 1000 simple calls with no program/texture switches is like 0.5ms for me.
lmao what the fuck are you using, JavaScript?

>>105948859
nvm 0.15ms, c++. 0.1ms for 100k DrawElements() isnt true. if you do indirect/instancing then yes. post a screenshot

>>105948848
"draw call" is meaningless. draw calls without state change just push the draw command in the internal command list.

>>105948848
Not anon, but I find that if your GPU is busy because your are uploading a lot of work to the GPU the time spent on the GPU scales based on work (the number of triangles matters more than the number of draw calls).
And there are like 100 different reasons for stalling and etc.
One example is that many tutorials have a wrapper to call glGetError for every call / Debug context is slow (you may want to use both glGetError + debug callbacks because without glGetError, the callback will be triggered at some later point).
Another reason is a sync point (modifying / streaming buffer data, it's really bad if you reuse one VBO for all models in a single frame).
But you would probably need to profile to know, and it's not so intuitive to read (and I will admit, one big annoyance with opengl is that it is not getting the latest updates in profilers / debuggers, you could use Google Angle if you don't mind reading DX/Vulkan profiling data since Opengl Can run ontop of those API's)

>>105948945
its not "meaningless". # of draw calls is a rough approximation of how much state changes you are issuing and how much driver overhead there will be.

>>105948909
My entire game which had a 2D UI + 3D game area ran its entire rendering phase in 0.3 milliseconds, no batched sprites or anything

>>105948968
>draw calls is a rough approximation of how much state changes you are issuing
A "draw call" can be 10 state changes or 0 state changes so it's a pretty bad estimate

>>105948968
if a draw does nothing then it does nothing

>>105949001
no? there is still driver overhead.
post a screenshot of you doing 100k DrawElements() calls in a loop in 0.1ms. I dont believe you.

>>105949073
I'm still looking for it

>>105949130
13ms for me. I have a hard time believing in a 130x slowdown but ok.

>>105949073
Actually i was wrong, it's closer to 10k in 0.1 milliseconds

>>105949073
he is lying since each call would cost 1 nano second.
but he already told you why, you just aren't listening.
he is saying the call is being queued up to the command list.
if he measured each call individually, it's not unusual for the function to only take around a few hundred nanoseconds to execute, since all it's doing is copying the command to be executed later.
and if anon has a perfectly zero sync point rendering system (allowing the buffering to utilize 100% of the GPU), it's possible that with a 60hz display with vsync, 100,000 draw calls takes less than 16ms, which means that the GPU is mostly waiting for frame to be drawn, so the actual drawcall itself takes no time at all.
You should use opengl Timers to measure the time, but that should take much longer than 0.1 or 0.5ms for a single draw call (you would need to have multiple timers to handle each buffered frame, usually just 2 timers is enough).

>>105943307 (OP)
a question all the AI fag ignore if its so good why can't the AI write a non shit CUDA alternative for amd/intel that allows their hardware to be used for training/inference?

>>105950016
Because AI can't create anything that it hasn't already been trained on. Zero innovation possible. Also because AI can't create anything too complex without completely shitting the bed. AI fags will attempt to refute this with zero evidence.

>>105950839
Two more weeks until AGI!
This time for sure.

Has anyone tried making a rule-based engine? I'm struggling with conflict resolution and event chains.

>>105948090
yes that is what I said and (You) can twist it as a runtime vs compile time thing to justify yourself to prove that statement of me as wrong but there is a reason why no compiler does the things the way you are arguing about and rely on v-tables instead for actually supporting methods.
The irony is that Odin, the language you are so arguing about, is probably the only closest implementation to what you are yapping with its "->" operator by not relying on a vtable. You can use parapoly for your type issues if you don't want to cry structs themselves not inherently supporting them. Truly hilarious.
>>105948321
>Anyway I'm using Odin if that makes a difference. Figured I'd learn a "modern" language.
here is a relevant resource that might come handy
https://marianpekar.com/blog/2d-cloth-simulation-in-odin-with-raylib
>are there any good explanations of the graphics pipeline
https://www.khronos.org/opengl/wiki/Rendering_Pipeline_Overview
if you are a beginner then stick to raylib and opengl for a month or two before diving into vulkan.
learnopengl.com is good for beginners to get an understanding of the traditional graphics pipeline. Modern graphics pipeline is significantly simpler and smaller but only modern graphics APIs like Vulkan, DirectX or Metal have solid support for it so leave that for later.
>>105948621
multi threading and manual memory management is more work because it gives more performance so no one actually complains about it, unless those who have skill issues, of course.

working on some weapons and stuff. lua is comfy to use when youve got it setup right. i have type hints and intellisense and it makes it pretty usable to do the game code in. hot reloading of scripts is so awesome.

>>105951493
>we got a /gedg/ control clone before gta 6

>>105951447
bro what the fuck are you talking about, it just sounds like random gibberish at this point
the point was Odin doesn't have methods, you can just admit this instead of typing out pages of weird cope

>>105951562
>the point was Odin doesn't have methods
I never said Odin has methods though. I got the trip, feel free to look up my previous messages to force an L on me.
Its the other anon thats been talking in circles but yeah, you can't do shit about them so take it on me instead lmao

>>105951584
You got mad cause I said Odin didn't have methods
Said they needed vtables, but they don't, I tried to explain this to you but you don't seem to get it

>>105951590
>Said they needed vtables
yes. Any production grade compiler has to rely on vtables to support methods. Only toy languages will attempt to support methods without a vtable and though you might see it like one, Odin is not a toy language. You kept on with your dumb take on "doesn't need vtables" but failed to provide any practical application in existence. My previous message was about Odin supporting methods, it was you who went on runtime vs compile time shit. You even said generics are not needed at one point and then changed it to generics are needed.
Its totally you bruh, you gotta try harder if you want to make it look like its me.

>>105951625
Methods can be dynamically dispatched or statically dispatched
Only dynamic dispatch requires vtables
Methods are just functions with an implicit first parameter

>dynamic dispatch
i seriously hope you guys don't do this...

>>105951632
talking in circles, are we
refer to >>105926841
>if that is your definition of methods then you should have no problem with methods in Odin because Odin has "->" operator for exactly this purpose. Stop crying over pointless shit like a retard already.
if that is your definition of methods then you should have no problem with methods in Odin because Odin has "->" operator for exactly this purpose. Stop crying over pointless shit like a retard already.

>>105951660
That operator does dynamic dispatch
Odin still doesn't have methods
This really isn't difficult to understand, a method is a function that has an implicit first parameter

>>105951674
an implicit first parameter that also serves as a context, I should add

i have absolutely 0 idea about game design aside from playing games, do you guys think a solo dev could do better than this piece of shit? https://www.youtube.com/watch?v=y3SRzgHm16M&t=260s

I love this sport and just want a fucking competent game

>>105951736
probably not

>>105951736
sports games are pretty good for new devs actually.
You have a fairly rigorous reference system already, so it'd just come down to if you want to tune it more arcade-y or realistic

>>105943307 (OP)
Can I use my Python knowledge in Unity? Or do I really have to learn C#? I don't wanna use Godot

>>105952039
Write Python -> C# transpiler.

>>105951736
the ai and animation seem like it would be pretty hard and thats the part that makes the game good.

Why is Lua bad? Is it because it's a scripting language?

>>105952482
its not bad
lacks features though

>>105952503
What features does it lack? Isn't it fairly aged by now?

>>105952612
no switch / pattern matching
no augmented assignment operators
would be better with an array and a map type instead of a combined type

wgpu is actually really cool

>>105951674
>>105951682
>this is what a method should be
>shows "->" operator
>NO A METHOD IS MUCH MORE THAN THAT
>goes with random bs as usual on the second run
At this point, even your AI assistant will be struggling to keep up with your bs lmao
That operator does not do dynamic dispatch. It can be considered as dynamic dispatch but no, its different because the implementation is not "dynamic dispatch" as in the traditional sense.
>Odin still doesn't have methods
I never said they did. Its your bs that is making all the unnecessary trolling here.
>a method is a function that has an implicit first parameter
this is literally what Odin's "->" operator does. Its designed to serve as syntax sugar for DirectX kinds COM APIs and Odin does not use vtables for this(or anything at all).
>an implicit first parameter that also serves as a context
this Odin does not do and this is exactly why vtables are needed. No language in production does this without relying on vtables, no matter how much you cry that vtables not being needed for methods, this one thing alone ensures that they do.

After using C, I see no reason to use ASM unless you're putting the hobby into hobbyist.

>>105953554
this is just vulkan 1.0 for javascript.
other than managed resources and barriers, it has all same problems of vulkan of being hard to make dynamic rendering.

Taking a year off work to spend time with my son do omscs computer graphics classes. I previously finished about half of it. Also want to build up my portfolio to get hired as a game or engine dev. What are some good portfolio projects? I also have a strong industry connection and they use Unreal. Tired of big tech and I only need to make $100-150k going forward and I live in Socal.

Is this just a better version of /agdg/

>>105954744
No.

>>105954308
>he is bottlenecked by c's switch-case semantics
http://lua-users.org/lists/lua-l/2011-02/msg00742.html

>>105944572
>deprecated on apple shit entirely
Who gives a fuck about Apple shit? I'm not making some Farmville clone for Facebook soccermoms.
>>105945057
People post this but nobody ever gives good reasons as to why.

Does it still use descriptors or similar outdated shit like Vulkan does?

Instead of complaining about LLM systems sucking at gamedev, you should design projects around what they can deliver, so your shit can radically scale in these areas and make a difference.

I'm a wretched asset+idea guy with some programmer buddies that I could call up and I'm currently preparing documentation for the next coding Grok/GPT model. The project is a clone of an old school game and is so simple, that I might not even have to call in collab to unfuck it after the AI is done with it, but the play time will be like 8+ hours of decent content, probably. I just have to build it around tacking on new event flag classes. The AI can notice narrative inconsistencies as well, so the bloat will be less demented-sounding. Find yourself something dumb enough to make it work for you, basically.

>>105955260
>you should design projects around what they can deliver
go back to /agdg/
/gedg/ mostly has people who build or configure tools that closely or directly interact with the hardware for performing computer graphics shit, meaning simple and complex simulations that can be used in a video game. Its rare to see a typical /gedg/ fag focus solely on making a game without building custom tools that go into making game themselves.
The "building custom tools" part is much more fun than finishing the development of a yet another slop clone that any AI can use as reference to generate a clone of that exact game without any copyright issues, making both (You) and (You)r game obsolete.
But AI will always struggle with "building custom tools"(a limitation of LLM) so we /gedg/ fags will outlive AI and chill without a worry unlike the /agdg/ fags.

>>105948129
yes one of the best sources honestly

>>105955435
>we just autismmax and do tools instead of whole projects!
>we're heckin better than AGDG!
My bad. But the moment you decide to make a finished game, what I wrote still applies.

>>105954333
No no I'm talking about the library. It uses the WebGPU spec to create a cross-platform graphics API.
>dynamic rendering
Yeah I don't think it has it.

>>105955744
it doesn't have it but you need it. that's where the pain of vulkan and co comes from.

>>105955581
>But the moment you decide to make a finished game, what I wrote still applies
good luck using AI for creating something awesome like terraria or factorio.
The most you can do with AI is another flappy bird clone so unless mediocrity, at most, is what you strive for, you are ngmi allying with AI.

>>105954731
>engine dev jobs
Do these exist?

>>105953778
you dont understand what you're talking about

>>105943307 (OP)
Before Grok existed we had this ancient technology called "git" that would allow you to download those ThreeJS examples directly. Shame that technology has been lost to time.

So... there really is no way to sort sprites in 3D based on their Z position rather than distance to camera in Godot?
WTF this is like the most basic feature you need for a 2.5D game
Thanks for wasting 2 days of my life Godot

is anyone actually productively using llm agents for developing a game? I don't mean bouncing ideas off or copypasting from chat, more something like claude code or some fully local agent?

>>105956801
No. I use Claude 4 Opus and it still has to be corrected constantly. Not really in terms of outputting code that doesn't work because 95% of the time it does (at least in C++), but in terms of system design etc it proposes overly complex solutions that are usually both worse in terms of performance and maintainability/complexity.

I'm pretty sure all the people you see online talking about using AI to that extent or AI replacing them are webshitters and jeets. AI isn't replacing proper programmers, AI is replacing the useless JS library slopper that uses IsEven.js and takes 3 months to make a webpage highschool kids in the 90s made in a weekend.

shut the fuck up about AI

>>105956897
That said
>>105956122
>The most you can do with AI is another flappy bird clone so unless mediocrity, at most, is what you strive for, you are ngmi allying with AI.
This is an absolute low IQ NGMI mindset, if you aren't using AI you're a total moron and you are already massively behind those that use it, and you will only be more and more behind. Programmers not using AI will be like metalworkers not using powered tools except they won't even have the "muh fully handmade" appeal since it's code, not an ornamental sword or something.

>>105956920
Using AI makes you a low IQ moron
You're a competitor at the special olympics

>>105956898
cope im going to animate a 3d model on vulkan with ai and shit all over your pop a bubble or move a sims house garbage

>>105956991
im real happy for you, you can go post about it over at /agdg/ and not the game programming thread

>>105957000
i said in vulkan

>>105957002
vulkan doesnt have anything to do with animation smoothbrain

>>105957018
oh yes it does

>>105957024
vulkan is a GPU API

idea guys won in the end...

>>105956948
there's dependency on ai and using ai to enhance your workflow

people who get dumbed down by using ai are people who never learned to program as an example

>>105957084
If you know how to program then AI isn't enchancing your workflow

>>105957117
i made ai write boiler plate code which is enough for me
it can help with mathematics for the most part when making algorithms

>>105943307 (OP)
>GR0K 4 EDITION
more like OP IS GAY EDITION
>only 300$/month
yo mama not worth it.
3.99$ max, take it or leave it.
>>105956920
>This is an absolute low IQ NGMI mindset
says the guy who actually can't do shit and needs to pay for a premium LLM model to get even basic shit done. The epitome of incompetency.
>Programmers not using AI will be like metalworkers not using powered tools
no, that's just your dumb analogy, showing you are an absolute idiot.
Metal workers using powered tools can craft tools with precision but programmers using AI can only do a basic shit like space shooter or flappy bird clone at most. Its programmers who do not use AI that create awesome, mind blowing shit.
>won't even have the "muh fully handmade" appeal since its code
only an artisan can truly experience the pleasure of handcrafting their work with their tools.
Incompetent piece of shits just don't care about art and shit over everything. Unlike what you stated, "muh fully handmade" code works seamlessly well with teams while AI slop doesn't.
Stop projected your incompetence under the disguise of using AI.

LLM discussion belongs on another bread. I blame OP. Fuck you OP.

>>105958039
non-retards need to make the OP
maybe I should start doing it again

>>105956792
I assume you figured out how to do this because this is a baked in feature.

>>105957145
You are pretty clearly too stupid to use AI properly and think using AI means creating flappy birds clones in JS.

I'm not going to correct you because reality will teach you, enjoy being a laggard and getting left behind.

>>105959687
>You are pretty clearly too stupid to use AI properly
the irony of this statement

>>105959690
He's a prompt artist/engineer. He's clearly smarter than you.

>>105958039
how is LLM not game development related??

>>105960386
fuck off

>>105960216
nice if you can actually simulate gears youd be onto something

not sure if i should have entities with a list of children or with a parent pointer ...

>>105960501
I have both

>>105960386
The same reason /gedg/ is segregated from other generals like /dpt/. "related" things that dominate other generals are put in containment threads. There are 5 containment threads on this on this board alone on the catalog as I speak, all dedicated to LLM discussions.

>>105960523
There really should be an AI board

>>105960596
Absolutely

>>105943307 (OP)
Do you ever completely lose it when someone questions your game design decisions?

>>105961536
50+ year old man with two successful games under his belt still acts like teenager who hasn't recieved enough validation from his peers

>>105960400
Thank Roblox for Augmented Vertex Block Descent

>>105944319
Honestly how much does it take to read this thing? Should I imagine maybe... six months?
Ugh, I feel like in six months AI will do all the programming. What's the fucking point

I've been toying around with Raylib and some rectangles.
However it feels like the only way to learn this is to read the examples on the website (and they even say this). But at the same time, it feels like cheating. Right now I'm trying to write a very basic platformer and I'm struggling to figure out how to calculate when the "player" is touching the "ground" or a "platform" and when it can jump and so on. Again, I am reading one of the examples in the website but it somehow feels like cheating.
Is there a better way or should I just keep pushing?
Should I just keep writing shitty code or do I try to draw everything on say, pen and paper first?
Any tips welcome.

>>105961656
90% of OpenGL commands are useless leftovers from the old days.
Modern GPU programming works like this
>Make canvas
>Make databuffer
>Fill databuffer with slop
>load program (shader) onto GPU
>program churns slop in the databuffer to create pixels on the canvas
>'flip' the canvas (give a pointer to your OS to hand off to a driver to shovel down the throat of your monitor)
>repeat

I looked at agdg once for 5 minutes and got ragebaited. nobody cares how anything works anymore it's just what asset you're supposed to buy on the unity store.

>>105961720
i hope you're not using SDL/GLFW

Is it worth going back to the basics on math? I went to hick school in the USA. I have a working knowledge of most math up to calculus and know a little LA. I was considering taking a couple hours a day and working through Basic Mathamatics -> How To Prove It -> Linear Algebra Done Right. Hard to tell if I know enough but a inferiority complex is there. I have about year to study as I am taking a sabbatical.

>>105961810
all the math you need is the ability to visualize a matrix transformation in your head

>>105961680
You have to make it so that if the object touches the platform that it is immediately redrawn where it should be
.t Just finished watching a Raylib Tetris tutorial

>>105961680
>cheating
You don't want to be the inventor. That's a long life of work for one answer. We live on the shoulder of giants for a reason. Use it.

Should I use SDL's thread or C++'s thread?

How did people deal with C dependencies back in the day? Including a C++ library that uses a single legacy C header is already enough to unleash a hell in my global namespace making the suggestion list borderline unusable.

>>105962389
Does SDL thread needed for SDL stuff? If either works then use whatever has the cleanest interface.

>>105962401
libname_functionname
also, only include headers in files that use them.
It's really not hard at all, I don't know how this confuses and enrages the modern programmer.

>>105962146
>>105961999
Thanks anons.
One other thing. Should I just stick to C for now or should I invest the needed time to properly learn C++?

>>105955054
>Who gives a fuck about Apple shit?
Anyone who likes money. Macs are 15% of the desktop market (even when you count the vast third world hordes with their potato PCs and worthless currencies, so their percentage in the west is probably way higher) and their users have lots of money to burn.

>>105962454
C is simple, stick to C. Yeah a lot of code you'll find are C++, but it's just class BS you can use a struct in place for.

>>105962440
That approach doesn't work for stuff like <inttypes.h>. Once you use is as indirect dependency in one of the headers it's getting hard to contain. Another example would be using GL types like GLenum which you sometimes use in the structure declarations expecting it would be the proper thing to do.

>>105962546
I guess I don't see why having inttypes available globally is so bad in the context of a program.
Still I get the point you're making generally, and usually what I do if I really *need* to separate some symbols from eachother is I write a little wrapper that exposes just what I need with the names I want them to have.
Then the rest of my program uses the thin wrapper.

>>105961680
>using a dictionary to learn new words in another language is cheating
bro

Progress:
- Updated animation manager keyframes to play sound effects. This way spells that do a "spell cast"->"spell moving"->"spell hit" effect at the given time properly.
- Sound manager had a limit for n-sounds that can play at a time via a pool. I adjusted this to also have another n-limit per a specific sound effect as well.
No webm audio but whatever

>>105959687
>I'm not going to correct you
you can't.
Feel free to prove me wrong by making AI generate a game that has several hours of progressing game play within one prompt. We all know how shit AI is for anything serious so keep on coping.
>>105960216
peak.
>>105961656
>Ugh, I feel like in six months AI will do all the programming. What's the fucking point
it was supposed to do that six months ago though. No point in worrying about that.
>>105961680
>I'm struggling to figure out how to calculate when the "player" is touching the "ground" or a "platform"
I usually rely on ray-casting for this. I draw a ray from player to ground and if the ray hits the ground, its a hit. Its just a simple (a+bt) math where a is ray origin(player's position), b is ray direction(downward in this case) and t is the distance(start from 0, snatch whatever is hit while ignoring any hits detected on player).
>Should I just keep writing shitty code
yes.
There is a difference between prototyping and production. You worry about not writing shitty code(the level of shittyness is decided by how much your co-workers approve of your code) when going into production but when experimenting, learning or protyping stuff, there is no such thing as shitty code. Its all about getting shit done so don't let it bother you.
>>105963073
still going stronk, eh!
noice!

>>105962466
>but it's just class BS you can use a struct in place for.

NTA but writing pointers to functions for use in a struct is a pain in the ass. Answer: make it .cpp but that's about as far as it goes though the STL does seem cool.

>>105962401
Dynamic linking

>>105963073
Very cool.

anyone tried using AI tools for audio and voiceovers? I'm kinda digging it, and there's no way I could even learn to compose songs, and even if I did I wouldn't have the time for it, so it seems like the obvious choice
same for voice acting but for different reasons, scouting, managing them, having to pay them without even knowing if the game will be a success
https://voca.ro/1eQtxwAuPeVG

>>105961553
jblow is a proper autist, the kind you'd find on 4chan, the OG kind that gave nerds a bad rep
i don't like him but you've gotta respect his integrity

>>105963524
>you've gotta respect his integrity
I don't because he doesn't have any
He routinely throws tantrums and pretends to know things he doesn't and blocks people when they point it out
He doesn't stand for anything, just another guy with a Napoleon complex

>>105963533
I don't like him but that's his personality. He's not trying to be someone else like every other streamer putting on a fake nice and cool persona when most of them are assholes and cunts. Maybe he's too autistic to fake a nice personality, idk, but in a sea of people faking their whole selves I appreciate someone keeping it real even if I don't like him.

here's my webgpu blog
>try to make deferred renderer
>hit color attachment limit
>set color attachment limit to be higher
>still hit it
>try to use smaller texture format sizes
>can't do it since webgpu is extremely picky about compute storage texture formats since a feature flag isn't implemented yet.
>move color attachments to storage textures that are done during the fragment shader stage
>depth isn't respected
>look for pre-fragment z test flag
>issue opened, low priority - for the spec. would take longer for actual implementation.
mfw

I think I'm just going to build the camera view gbuffers by using color attachments and just doing the vertex assembly multiple times.
It's a bit stupid but I've had enough.

>webgpu
seems like a giant meme

>>105943307 (OP)
Should I learn Phyton then move to GDscript/Godot later when I'm proficient with it, or just dive straight to Godot?

>>105964088
Dive straight into Godot.

i wish i just implmented lua earlier god damnit.
im cutting out so much "data driven" fucking bullshit in my code that was a work around for not having a fast iteration language.

>>105951257
Have you seen this:
https://www.youtube.com/watch?v=tAbBID3N64A
and/or have you tried using prolog?

>>105964621
isn't this against WB's nemesis system patent copyring?

>>105963592
>2025
>deferred renderer

>>105964657
Left 4 dead 2 came out in 2009

>>105964681
that's wild

>>105964677
clustered deferred > all

>>105950016
It kinda does for me. I use gemini CLI to backport scripts and functionality from windows to Linux since I jumped ship from w10 to arch. Including some llm inference stuff.

Honestly I think the big limitation is people not trying to do things with llms assuming it doesn't work or "someone else must have done it by now". Genuinely just try doing it yourself you'll be surprised how far it gets. If it makes a mistake point it in the right direction.

>>105963286
The way to do it would be writing a function that takes the struct as input, or better yet an array of structs.

>>105964790
>It kinda does for me. I use gemini CLI to backport scripts and functionality from windows to Linux since I jumped ship from w10 to arch. Including some llm inference stuff.
your comment makes no sense how does the LLM backport script that somehow fix the mess AMD made?

>>105964790
>>105964823
wrong thread guys

how do we LOBBY effectively for an AI board?

If there was a game dev board, we would be kings.

>>105964875
>getting 4chan to do anything
dude, hiroshimoot wouldn't even update linux to patch 20 years of security issues until after the website got hacked

>>105964153
won't he learn theory more slowly, if he's not solving problems written in a widely used language?

>>105965594
Honestly the Python and GDScript comparison are so surface level, I think people only do it to pull in other developers easier into the Godot ecosystem. GDScript is a widely used language anyways. More fitted for game development than Python since GDScript is a whole library of game development related functions and it all works in the Godot editor that lets you monitor all the things you would want like objects, performance and a convenient console window.

Is there a point in using OpenGL other than latest (4.6)? Aside from self-imposed limitation to make sure you don't go past your target graphical style. Are older OGL versions simpler and easier to get going in? I would imagine that for the same result, OGL4.6 should do things at least as easily and as fast as OGL2, likely faster and easier.

>>105963592
>he fell for the webgpu meme

I could mock WebGPU for all its' faults, providing you nigs free talking points. Orrr, orrr, I could leave it at "ahaha they fell for yet another poverty abstraction library. One made by and for incompetent webshits that still havent agreed on anything no less"

>>105962565
Turns out you can #include inside a namespace, but either you need to be consistent with it or it doesn't always work. Won't help with #define though.
Thin wrappers are the way to go in general.

>>105948968
>more than O(1) state changers per frame

>>105962460
>Macs are 15% of the desktop market
And 0% of the desktop gaming market; even Linux has a higher market share.

>>105951736
Sakurai says that sports games are sort of dubious because the appeal is "I want to play as my favorite baseball player," which would mean piles of licensing, which would make it unreasonable for an indie guy. But something less realistic would work.

>>105952482
No typing, standard library is a little under-equipped (I have to write my own table print function? Are you for real, my guy?), no continue statement, globals by default, 1-cringed indexing. But honestly none of that is a deal-breaker. The upside is that it's well-known, not totally insane, you can run the scripts directly, and it's easy to embed. JavaScript would probably be a better language to use assuming it's as easy to embed as lua, but maybe it en't.

>>105960216
NORMALS GET OUT REEEEEEEEEEEEEEEEEEEEEE

>>105963073
>>>/wsg/ has 'for other boards' thread.

>>105964875
Getting a degeneracy containment board is probably too much of an ask considering the state of most boards I look at.

>>105962460
>And 0% of the desktop gaming market; even Linux has a higher market share.
I'm just making that up DESU but it feels true.

>>105963592
>>105966050
Honestly I'll never use WebGPU due to the name alone, always makes me think of it just being made for webjeets writing shitty non-games in some JS framework.

>>105962460
15% doesn't matter then 14 of those points are SanFran hipsters with a MacBook sitting at Starbucks writing their opinion piece on why drag queen story hour attendance should be mandatory in kindergarten.

I'm not making a gay dating sim, or a game for women, they'd likely be way way less than the 1% in pic related for my specific case.

>>105965854
The highest version of opengl es you can use with emscripten / WebGL is GL es 3.0.
Angle is nice, you probably don't have the autism to care about it, but it seems like desktop opengl is going to use an older way of composing the contents onto the screen using "Copy with GPU GDI" shown with intel presentMon (you need to add the compose/present method widget to the overlay). DX11+ added a new presentation method called "flip" I think, and with the new method, you can alt+tab and toggle desktop fullscreen without the screen flickering for half a second, and while in fullscreen certain overlays wont work without flip mode like adjusting volume using your keyboard / IME text input (emoji menu, Win + . ), but the steam overlay will work fine with opengl/vulkan.
On an nvidia control panel you can set Vulkan/Opengl present method to DXGI and it will make opengl/vulkan behave the same as DX11 (3d settings, at the bottom).
Also win 11 vs win 10 behaves differently due to fullscreen optimizations (I never used windows 10, but I use windows 11 so I don't really know what it is, All I know about fullscreen optimizations is that if you force it off in the executable compatibility settings, it allows your application to use exclusive fullscreen which is when your entire screen gets resized and freezes for 1-2 seconds and I remember in the past it wont resize itself back if the program crashed, which is not the same as the black screen flicker I mentioned for alt tabbing/fullscreen toggle, it's a lot shorter and minor, I also think if I disable Vsync and the fps is above 500fps, flip mode has a significantly higher fps compared to desktop fullscreen I think, BUT exclusive fullscreen+disable fullscreen optimizations has the same result).

is there an underlying technical problem with rpgs having fake choice and being super-linear, while pretending they're open-ended? Shouldn't this be solved and some best practices be in place for forking story lines etc?

>>105967673
Correction*
This might not have anything to do with flip mode, the flicker / overlay issue might be only from DXGI.
Angle with DX11 also shows "Copy with GPU GDI"
I tested flip mode with EGL_ANGLE_direct_composition, and I think it gave the FPS boost, but that's about it.

>>105967712
There's an underlying economic problem
Forking the story means you have to make double the content, and most people are going to see only one path, it's a massive waste of resources

After 6 hours, I finally got SDL3 working on the web, with source-level debugging and all.

>>105967712
Detroit Become Human did a good job showing the branching paths of each level. Modern RPG's could never do that because it would just be a straight line.

>>105964621
>https://www.youtube.com/watch?v=tAbBID3N64A
This is great.
>and/or have you tried using prolog?
Prolog works for my needs but I'm using Python with sympy since I'm still experimenting.
Currently, I enumerate all valid solutions for all rules, and use a variation of the Rete Algorithm to build a DFA (instead of a graph).
The problem is that many events conflict with each other (event "A" modifies a variable after event "B" reads it, possibly leading to a corrupt state). But you might want to allow this for event chains (event "A" changes a variable, "B" is now valid so we trigger "B" as well).
Since the correct behaviour depends on the use case, I'm going to implement different conflict-resolution strategies and let the developer choose.

>>105935661
> transform matrices are arcane magic
are you using raymath? Then you can blame that because it sucks and is really only intended as a quick and dirty solution.

>>105968167
Nice. But...
What programming language is that?
>.bf
>probably not brainfuck
>case .THING isn't C?
>Console.WriteLine("{}") sounds like C#...
>let var = .(); ???
>not Zig, not Odin...

>>105969862
brainfuck

>>105969862
beef

>>105964621
>PCA to cluster your rules in a partition
lmao

>>105963073
1/2

>>105970678

If you want to learn programming, make your own game engine
If you want to make a game, use an existing game engine
If you want to make a great game, make your own game engine

>>105971334
no, the great game part statistically depends more on working on a team than "making a game engine". You are just doing A, and then you are doing B at the same time, and you need to pick which part you are going to prioritize (unless it's a 2D game, the engine part doesn't really exist in 2D games, it can sort of merge with the game code where it's hard to know which part is engine and which part is game.
If you work alone, you just simply don't have the time to make a game in 1 year that matches the quality of a game that a team of artists / developers that are masters of their own domain can potentially make.

>>105971778
Nah, I'd win.

Any insights on how to store a shmup level effectively?
Planning an on-rails shooter(basically the 3d version of that) and so far the idea is just a list of things that spawn(with a path and whatnot) and at what times they come, but that feels naive somewhat.

>https://www.youtube.com/watch?v=2vnTtiLrV1w
>stb_voxel_render.h
>"This was probably 3.3-ish, but I might well have been using 1.1 + extensions."
what's your excuse? where's your efficient voxel engine?

why arent you idiots working as a team under and following the orders of the ppl that know more than you(meritocracy)

>>105972032
that's called a hierarchy not a meritocracy and why would I follow the orders of somebody who isn't paying me

>>105972082
because you guys are not making a game engine otherwise.
>a meritocracy isnt hierarchy
yes it is. What are you an anarchist.

>>105972099
meritocracy and hierarchy don't have a correlation
why would anyone want to make a game engine for somebody else for free?

>>105972099
is a rectangle a square

>>105972169
no a square equal on all sides

>>105971778
Too true xister, animal well, stardew valley, noita all fucking SUCK!!!!!

>>105972233
I said 1 year and noita isn't made by a lone wolf developer.
Also stardew and animal well were made with a publisher late in the game development (not to take away from the programming / art skill of those games, but I think marketing is an essential skill for profit maxxing. I think those games could have done fine without a publisher, but we wouldn't know)

>>105972007
I used to think voxel would be the future, but that kind of changed with AI. Now voxel is just a maths problem that no one may be solving anymore.

>>105972490
voxels are less efficient than polygons

>>105972539
And yet capable of much more than polygons could ever be.

>>105973017
That doesn't mean anything

>>105973041
There’s a reason digging and destruction are best done with voxels and not polygons.

>>105973208
voxels are good for physical representations of space
polygons are good for rendering ie. what you can actually see

wtf is this discussion? you guys are pretending that those things are mutually exclusive when in reality they rely on each other. destructive terrain uses marching cubes for rendering anyway

>>105973245
you can render voxels but it's a generally a dumb idea

While the vibe coders are copy/pasting, I am actually doing pixelshit. Building the missile, testing the missile, racing the missile. The missile always knows where it is. And always knows where you are not. Instead of memes and hiding behind PRNG. Get your bitch asses out of the chatbots canvASS and progress post something.

got an idea on how to implement extrude region

>>105973270
what in tarnation

I got grok3 mini-fast (free) to make the old SkiFree clone, even had it generate background music, speed boosting music (faster), death sound and the yeti chase. Took a while tho, since I fucked up a bit earlier by asking it to do a 3D remake which generated 3D game but the scene was fucked up, and asked it to do 2D

>>105973481
wrong thread

>>105973481
show it in motion

>>105973581
it's in the thread title moron, grok 4 is the state of the art for dev work and the coding model isn't even released yet.

>>105973934
nobody cares what you erroneously titled the thread

lua with intellisense/type hints is so cracked on god frfr. i cant believe i fell for the json data driven meme, just write it in lua.

>>105968167
holy fuck
@grok is this true?

>>105973964
You can use both you know

>>105965594
no. I'd suggest to dive straight into GDScript as well.
Programming is about problem solving, not language learning and GDScript lets you solve video game problems more effortlessly than python. Learning GDScript also makes learning python easier
>>105969862
its beef.
It would have been widely adapted if they marketed it as a C++ killer.
Shame they didn't and those who accidentally tried it looking for its OOP-ness ended up appreciating it.
>>105971334
trvth nuke

>>105974027
true, i use json still for the world editor and import settings. but desu anything else is better in lua. a programming language is the ultimate data driven format. you can use shared constants, use functions etc.

>>105972539
>>105973041
the only valid use case I found for voxels is visualizing MRI scans.
Much efficeint than polygons for this use case.
>>105973301
If you need an idea, I suggest picking a computational geometry book.
You let the user define two vertices. When the first one is defined, get the adjacent edge and now you have three vertices at your disposal. When the second vertex is define, generate a quad from previous three and the newly provided vertices. Blender does some fancy, advanced nearest distance calculations to make this experience more seamless.

How do you get started on larger projects?
I'm trying to develop my first proper game without an engine. (just a personal preference, I think it's more fun.)
I initially tried doing a sort of "Split things into smaller parts as I go" approach, which just led to about 70% of my time being dedicated to debugging. It's pretty much a masterclass on how to create as much technical debt as possible.
How do you even get started without anything implemented to actually prototype anything?

If I don't get any more experienced ideas, I think I'm just going to start with conceptualizing as many "class"es as I can, and just seeing which ones need the least amount of dependence on the others, then working on those first.
Like, one to create the window, and display things to the screen, then one for each type of render-able object, input handling, file management, time keeping, game flow, etc.

I don't know why I freeze up so much when it comes to this. It's not like I'm scared, I just like doing things right, and I haven't taken any classes that teach more than the bare essentials of knowing what keywords mean, and how they work.

>>105974084
...you program in json? tom, is that you?

>>105975396
It's easy. Make a few games before so I know what I'll need and what I don't need.

>>105973017
Ok but the density of voxels required to make proper use of them outside of a voxel game is prohibitive even on the best consumer hardware available.

At best, voxels are just the underlying data structure that you then render normal polygons onto.

>>105975396
it's not fun.

>Juan 'reduz' Linietsky and Ariel 'punto' Manzur co-founded Codenix in 1999, a game development consulting company. As early as 2001, they began work on an engine then code-named "Larvotor"
godot is being developed for 25 years and its barely usable.
people just don't understand how complex game engines are.

>>105975981
An engine does not take 25 years to build, that's highly unusual

>>105975990
This. Godot is built by incompetent grifters whose collective experience building engines is, uh, waiting for somebody to open a PR with some poverty attempt at a feature AAA have depended on for years, just to end up closing it as "wont implement".

imagine:
>not having control over the loading of materials and effect/technique collections
>not having gi that isnt a single probe
(might've been improved by now, but still, no real gi. they're still behind decades old radiosity AHD solutions.)
>not having material layering of core world/nonmodel surfaces
>not having control over normal hardness
>not having any terrain building tools to give depth to a level
>not having any tess of artist defined control points like, i dont know, every game after doom2
>not having any real offline asset processing pipeline (pipeline resource compression)
>not having any real offline vertex input stream compression (pos, normal, bitan, tan compression)
>utter shit decal projection
>no world destruction (no glass for example)
>no real AI
>no real time world updates for a mostly dynamic game engine
>needing a supercomputer to run a game that looks like somebody's opengl 1.2 hello world game
>spending 1-2 decades sperging over how great your engine is, just to refine your argument down to "ACCEPT MY SHIT SCRIPTING LANGUAGE YOURE PROBABLY GOING TO REPLACE ANYWAY, OTHERWISE MY ENGINE HAS NO VALUE"

What's with this trend of "game engines" of no utility past being basic model viewers with physics collision?

>>105976095
They can get away with it because it's FOSS

>>105976095
What >>105976114 said, FOSStards will gladly eat up inferior products as long as they're FOSS.

>>105975990
proffesional game engines are developed by teams of dozens, if not hunders, highly skilled developers.
godot is what happens when a team of 2 mediocre programmers decide to make an engine. keep in mind that's 2x bigger than your one-man team.

>>105976196
or look at cherno's game engine. it's being developed for like 5 years, and it's in unusable alpha prototype state.

>>105976196
>proffesional game engines are developed by teams of dozens, if not hunders, highly skilled developers.
not exactly
most of the core engine work is done by a small team

>>105976207
Cherno project hops like no tomorrow, so it's fine. He teaches well and that's all that matters.

>>105965080
I messaged [spoiler]moot[/spoiler] asking for one [spoiler]10 years ago[/spoiler]

>>105976207
>cherno's
is a grifter. When will you faggots stop falling for "EX<COMPANY NAME> DEVLEOPER" tripping over the basics? No, they aren't coming to save the day with their brand new studio/engine/concept art/game. Especially not a braindead zoomer pandering to social media algorithms.

>>105976663
>grifter
>has one of the best opengl video learning series

>>105976095
>basic model viewers with physics collision
good enough 90% of the time desu

>>105976095
>not having control over the loading of materials and effect/technique collections
You can load whenever you want. You can even preload.
>not having gi that isnt a single probe
>(might've been improved by now, but still, no real gi. they're still behind decades old radiosity AHD solutions.)
There's been multiple probes for a while. Juan worked on GI for the past few years.
>not having material layering of core world/nonmodel surfaces
What does this mean?
>not having control over normal hardness
There's plenty of control, you can pull the values yourself and do anything.
>not having any terrain building tools to give depth to a level
Built in now. Assets are easily grabbed like Unity anyways.
>not having any tess of artist defined control points like, i dont know, every game after doom2
>not having any real offline asset processing pipeline (pipeline resource compression)
>not having any real offline vertex input stream compression (pos, normal, bitan, tan compression)
What? You don't need any of this.
>utter shit decal projection
It works doesn't it? You can do so much more on your own. There's no engine where the decal is perfect on its own.
>no world destruction (no glass for example)
What does this even mean?
>no real AI
What?
>no real time world updates for a mostly dynamic game engine
Huh?
>needing a supercomputer to run a game that looks like somebody's opengl 1.2 hello world game
Godot is the lowest foot print of Unity, Unreal and Godot. The only reason it would run intensively is if the dev was bad at optimizing. A common problem with having something like grass.
>spending 1-2 decades sperging over how great your engine is
This didn't happen. Godot is presented as an alternative and always has been. If anything, Godot is great simply by virtue of being funded by a lot of big fish in the pond.

>>105960216
edge extrude region(without duplicating shared vertices)
gotta figure out the triangle normals now

>>105977961
anon your windings

>>105977084
Here king, you dropped your crown. Next time you could leave it at "i'm a moron who cant comprehend a single one of your point".

>There's been multiple probes for a while. Juan worked on GI for the past few years.
bahaha

>There's plenty of control, you can pull the values yourself and do anything.
https://github.com/godotengine/godot-proposals/issues/1682
https://github.com/godotengine/godot/issues/18276

>Built in now. Assets are easily grabbed like Unity anyways.
Nobodys talking about an asset store or asset distribution, you fucking mong.

>>not having material layering of core world/nonmodel surfaces
>What does this mean?
>>utter shit decal projection
>It works doesn't it? You can do so much more on your own. There's no engine where the decal is perfect on its own.
>>no world destruction (no glass for example)
>What does this even mean?
>>no real AI
>What?
>>no real time world updates for a mostly dynamic game engine
>Huh?
you're a moron

>>not having control over the loading of materials and effect/technique collections
>You can load whenever you want. You can even preload.
if you say so...


Average godot tranny. They don't even realize they're literal decades behind where idtech derived engines were in the mid 2000s. Not to mention the proprietary shit that's been built up since the 90s.

>>105976095
some of these are unfair, but the rest are true.

the problem with godot is that its monstrously complex (and slow) (and dont hit me with that "simple" bullshit), and yet the engine gives you nothing out of the box other than rendering a model. it tries to handle everybodys usecase from 2d to 3d and it falls flat at doing anything well.
compare it to unreal/source/quake. they are engines specifically tailored to make 1 type of game (multiplayer fps/action). so you end up with an actual engine to make a game.
desu i dont see any reason to use godot at all. for 2d, a framework is easier. for 3d, unreal/unity make more sense. and if you want custom, its unironically easier to roll your own renderer instead of working around godots bullshit.

>>105975396
Rewrite the whole codebase and many times as you need. You accumulate code debt when you ignore obvious issues because your solution is "good enough" and "just works".
And write tests.

https://www.youtube.com/@BetterSoftwareConference
Hope you anons are watching these talks. Except the Muratori one, just watch the Thief 1 thing he mentions and save 2 hours.

>>105979812
this whole movement is kinda cringe because of Casey and his anti-OOP retardation

I am looking to create a game that has some FMV videos and some 2d based segments. What graphics library would best suit me here? I wanted to write it in SDL but there doesn't seem to be a native video playback option.

>>105980193
>he's wrong because ....

>>105981778
>I'm insecure please argue with me

>>105981854
>OOP is good because...

>>105981724
If you want HD video, you could use the chromium embedded framework. I forgot the exact details but I'm pretty sure you could find a tutorial with SDL + cef example (and you might use Cef for UI). This is still very difficult to get working however (maybe there is a new web library that's better than Cef).
If you don't mind engines, I think unity/godot has it built in, this is the easiest option.
If you want to make it look shitty, low resolution, and make the videos short (like 5 seconds), you could use gif, using stb_image + manually sync the audio (it's gonna use like 100 megabytes of ram when you load it, which isn't that much). If your video output does not support gif, ffmpeg or an online website can convert it. This still isn't easy. There is also libgif if stb_image is unable to play it (stb_image has given me a stack overflow error playing a gif) but libgif is a strange library, I had to make a few changes to get it working on windows (it does work).

>>105981868
*libgif
sorry, it's giflib

>>105981861
I've literally never seen someone who can be swayed on this topic by logical arguments

>>105981940
i would be very interested in what you think is a logical argument that isn't some appeal to authority/tradition

>>105981964
Well first you need to define what OOP is before you can say if it's good or bad
Is cargo cult programming bad? Yes
Is grouping data types and functions that operate on them bad? No, it's a very natural way to organize things. Casey's argument is "well sometimes it isn't the right abstraction" no shit, nobody says you have to do it everywhere

>>105982059
>Well first you need to define what OOP is before you can say if it's good or bad
and we're already done

>>105982064
refer back to >>105981940
I gave a logical argument and you ignored it and just replied to the first part
Waste of fucking time

>>105982074
>oop is good because it can very vaguely and broadly be defined [away from the relevant context where it's bad]
This isn't logical, it's disingenuous, this is why people hate OOP evangelists.

real oop has never been tried

>>105982121
Did you even read my post? That's exactly what I addressed, because OOP can be vaguely defined you need to specifically define it to have a meaningful discussion about it
Some things that can be called OOP are bad, some things that can be called OOP are good
Casey says the good stuff is bad too, namely functions bound to data types which what I consider the core of what OOP is

>>105982135
casey and everybody else already define the relevant version, you're either purposely ignoring it, or unable to grasp it. so again
>>>this is why people hate OOP evangelists.
but yeah i cba so don't bother responding again.

>>105982191
For the second time read, my fucking posts, it's not hard, they're not long. I addressed both of Casey's points. Yes the cargo cult is bad, but every style of programming has a cargo cult. There's a functional cargo cult, there's an ECS cargo cult, and there's a procedural cargo cult which he is partially responsible for
The technical features of OOP aren't bad, they're only bad if you use them in places where they don't belong
> i cba so don't bother responding again.
If you're too stupid to actually have a rational discussion don't ask
You're only proving me right

>>105981868
>If you want HD video, you could use the chromium embedded framework
holy shit zoomers are retarded haha

>>105981724
SDL + ffmpeg.

>>105943307 (OP)
can anyone recommend resources on making hand keyed 3D anims look at least semi-decent?

i made some without fully committing to blender/maya, just for prototyping, and i feel like i'd need to waste a lot of time fixing them because i dont really know how to animate

>>105981868
>chromium embedded framework

>>105964088
learn the one you will actually use

>>105965594
they aren't exactly the same. that's a meme.

models in every game I make are a bunch of cubes because I don't feel like learning blender

all you need are cubes desu

Dealing with issues related to my shitty coding, but I at least got torques and forces kind of working.

>>105981778
>OOP is LE BAD because computers 30 years ago had 8mb of memory
casey is such a dumbass

>>105983420
W so big its spinning in space

>>105983420
use grok 4 for debugging