Thread 106120386

YGG debut Edition!!!

/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/

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.

Previsous: >>106075373

I did not bake the bread but I welcome it openly, when is the next demo day?

Can I game dev in Python?

>>106121615
I have never explored 3d but I do fine with 2d in pyglet

>>106121615
Yes, for 2D you can use something like PyGame or Arcade. For 3D I've used Pyglet and made a few things like minecraft type block building prototypes. People will bitch about Python's performance but I like to use it just to churn out prototypes faster. And that's because the vast majority of the prototypes I make I'll just end up tossing. If you're going to the trouble of making your own engine you may as well experiment.

I've been out of the gamedev loop for about a year now. Any anons here still using Odin? Still enjoying it, or did it crash and burn for whatever reason?

>>106122331
>Still enjoying it, or did it crash and burn for whatever reason?
Odin hasn't changed much within a year.
If you didn't like it then you probably won't like it now.

>>106120386 (OP)
Best AI game dev: https://grok.com
10 games made with grok 4 heavy (only 300$/month)

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

>>106122516
>Best AI game dev
Prove that grok 4 is the best.
Enlighten me on whatever metrics you used to term it as the "Best AI" than the rest.

>>106122516
I've used things like Claude Code and Cursor on projects before. It's great at the start, especially if what you want is something that has a lot of examples on the internet. But once your project grows in size it starts to have difficulties. And I don't just mean like it doesn't just do what you want it do but all of these LLMs really like to generate code. Which is great when you start out but as the codebase grows it just becomes a mess.

From what I've learned is that these tools are fine to prototype in. But at some point using something like Cursor and hitting accept casually is only going to fuck yourself over. You have to learn to say right, from this point I'll just use it more sparingly, on very specific things (that are tiny in scope). If you ask it to add new features it just becomes unwieldy.

That's the hype part, where they're promising people you can build full and complex projects but notice two things: it's only like really short clips. And they never show you things like the codebase or even git commit history.

I say all this while remaining optimistic about the longer term. Like at some point we may have much better assistants that can actually handle a large project and not let the codebase go to shit but right now nobody has it.

For the items I have the context of the "right" vs "left" hand and the individual items on each hand. It works great for normal weaponry, but the bow (when shooting to the right) is the only weird one. Ignore the bowstring not moving for now.

>>106122516
rubic cube is nice if it scales rest is not very interesting

Working on rolls and backstep

>>106121559
>when is the next demo day?
its whenever you make progress.
Unfortunately, I don't think stuff like that makes sense for /gedg/ where learning how shit works under the hood is of the most priority and not shipping a game ASAP.
Everyone has their own pace of learning things and introducing a generalized deadline can be demotivating when the deadline was not met or it might also lead to rushed results with little to no understanding on what was worked on.

I added some sounds for archery https://i.4cdn.org/wsg/1754199820285048.webm For the sfx player I can set multiple sounds for a given key, this way some actions will cycle through a few different combinations (example: when youre mining a rock it wont always do the same clink)

>>106123625
I'll keep an eye on /v/ thread then, I thought the do something every now and then

>>106122537
told you once already, elon has the most gpus of any company, including 550k GB200s going up very soon just in time for the coding model. the frontier models are only as good as how much compute they use. grok super heavy uses the most compute power out of anyone.

>>106122690
this is correct, anyone doubting ai is poor or stupid. not even grok super heavy will one shot something complex, the point is to use it for learning and boilerplate garbage, and debugging c++ logic/syntax. you still have to put in the work. AI just allows you to use your own brain bandwidth on more important shit.

I am currently working on a single-player fantasy, first person RPG, that will also double as one of the most realistic melee combat simulations I have ever seen. Your sword will interact with your opponent's own weapon, allowing for the abilities to parry and deflect. You will only do damage if your sword touches an unarmored area of your opponent. Currently, the player can slash left/right, chop downwards, stab, and then pick up/drop items from both their left and right hands.
My next steps are to:
> Add both a horizontal and vertical block
> Begin Skeleton NPC
> Begin sdding shields

>>106123818
>I thought the do something every now and then
yeah, that's beyond me. Another trip fag Frosch's deals with that shit.
>>106123926
>grok super heavy uses the most compute power out of anyone.
I also told you once already that expensive does not mean better.
Just because Elon Mollosk has lots of GPUs does mean grok 4 better. Deepseek has already proved that point. Google, Microsoft and Meta have huge GPUs at their disposal as well, Elon isn't the only one with exclusive GPUs.
Your arguments of "this car has the most horsepower so its the best" is not valid because shit like torque, weight and gear ratios do matter a lot as well. Pointless paid shills.

>>106124278
*doesn't mean grok 4 is better

>>106122537
Just don't reply to the AI shitters

>>106123939
I can’t believe this isn’t more common. Keep it up buddy. A lot of potential.

>>106124830
yas saar

Odin bad

>>106125879
please explain why, is it just to be contrarian or what?

>>106125942
I'm just trying to trigger a tripfag meltdown

>>106126015
aw man imageboard schizos are the best wish I was here for it, haven't checked /g/ in a couple of months

meme languages will never be good

have gedge made a game yet

>been trying Jai
>the language is pretty good
>the metaprogramming is amazing
>I keep finding new things I can just metaprogram away
I kneel

>>106125879
then find a good language for yourself
>>106126057
then stick to your real programming language
>>106126753
good for you

>>106126584
keeping games aside, gedg made a lot of engines.
OP picrel is one of them.

bumping with progress. implemented basic textures

>>106128509
Damn, that looks really nice anon. What kind of game are you making and what are you using to make it?

>>106128981
thanks, i appreciate it, given how basic it is at the moment
>What kind of game are you making
blocking out story and puzzle ideas for a retro inspired sci fi survival horror game. got a few disconnected ideas in the making
>what are you using to make it?
c++ with directx 12

Tracklaying progress.

Moved from UE with OOP to straight C. Might move to something else later but I've only ever known OOP so I thought I'd try C first. What's the standard method of creating something like an entity type? In OOP I'd make a class, add members, set the defaults, and that's it, but that's OOP.

I can make a struct, then some sort of initializer function I guess? Should I make one for each "entity type" struct I have?

>>106130024
Yes, that is how it is.

>>106130024
struct and initializer function or just straight up struct declaration

>>106130024
When I messed around with Odin (similar to C) I made a struct with whatever defaults I want, a union for different entity types, and a enum to track the entity type stored. Though I think I was just re-inventing OOP at that point..

>>106130120
Got it, initializers it is.
>>106130144
>>106130195
>struct declaration
>defaults
You mean doing something like
struct Struct
{
int x;
} Struct_default = {y};

//then
Struct instance = Struct_default;

Because AFAIK there's no setting defaults inside the struct declaration in C.

>>106127232
engines are useless and don't work without games

>>106129097
train game ?

>>106130024
Unironically watch the first 30 episodes of Handmade Hero and you'll learn a ton, congratz on moving to C.

>>106130728
go back to /agdg/ and focus on finishing your game.
Just because nobody uses cry engine to make games doesn't mean that cry engine is useless. Its the other way around.
Games can be literally useless(as in cannot be used) and forget about working, they wouldn't even exist without an engine. All /agdg/ fags rely on engines to make games and struggle a lot with just that. Having balls to make engines is beyond them so mock /gedg/ fags with "where's the game" to self satisfy their inferiority complex lmao
Its retarded of you to assume that game dev and game engine dev are the same shit and trash talk game engines. an AI can make a game but it can't do shit when it comes to making a game engine. Your ignorance is totally off the roof. Keep coping with your "progress"

>>106127214
>mentioned Jai in the same post I shat on Odin last thread
>this guy's gonna be passive aggressive about Jai for the next week
kek

>>106131233
i might if you bait wasn't that low quality. Better luck next time.

Spent 2 weeks trying to improve performance.
Got like 10 fps in total.

It really is on me though, I should have profiled this shit, instead of blindly jumping into optimizing the first thing that came to mind.

>>106131251
Imagine if you spent all the time you did defending Odin actually making something

>>106131351
>I should have profiled this shit, instead of blindly jumping into optimizing the first thing that came to mind.
at least now you know what you should've done and that's still a W
>>106131377
>spent all the time you did defending Odin
that's just me replying in-between work hours. Can't work on making anything during that time and right now is me awake at 4 in the morning due to my insomnia.
>actually making something
I'm making a lot of things but they are not game dev exclusive shit so can't post any "progress" here. I also don't have a need for motivation or force myself to "make something" and seek acknowledgement from unknown strangers to up my skills. I just work on whatever I'm curious about and that's pretty much it.
I might post later this year after I started working on my first engine but that's no time soon. Still struggling with a path tracer and a decent game object system.

>>106131098
>go back to /agdg/
keep it up and I'll be linking your gay little club here in /agdg/ forever.

>>106131539
I can't stop you from doing whatever it is that you are thinking of doing so there is no point in me bothering about that.

>>106130355
Ah sorry, by defaults I just meant whatever variables I want to use for all the entity types. Default was bad wording on my part.

>>106131497
>at least now you know what you should've done and that's still a W
thanks, anon
well at least my thing is faster than godot, I guess.

Programming languages should have a keyword to use the last variable you used. Kind of like how he/she/it refers to the last person/thing mentioned.
So you could have for example
if(some_variable_x < 0) IT=0;
To set "some_variable_x" to 0.
Could even have it chain with IT1, IT2, IT3, etc to reference older accessed variables. Maybe get really fancy and IT() would be last function used

>>106132162
outside a few very specific cases, it sounds like a good way to make your code into an (even more) unreadable mess

>>106132162
its not that hard to implement, actually. Just poppin' whatever's on top of the stack but the code becomes very vague within no time, to the point that even you, the one who suggested, might get very annoyed by it while reading other people's code.

>>106132162
You can already kind of do this in assembly because you know what’s in the register.

>>106132162
seems kind of brittle
if you changed < 0 to < variable then it would change the value of it

>>106131098
> nobody uses cry engine to make games
remind please what engine far cry made with
what games use your engine?
if an engine has no single game using it such engine doesn't work and doesn't exist

>>106123371
>background is isometric
>sprites are flat
anon explain.

>>106132895
>what games use your engine?
I didn't even start working on my own engine yet though
>if an engine has no single game using it such engine doesn't work and doesn't exist
It takes a real engineers to accurately evaluate if an engine has proper support for good 3d rendering with flexible shader coding, physics, sound, along with other essential sub systems and how well all of them are integrated.
Using "games" as a metric to decide the potential of an engine is what the inexperienced who don't know shit do.
It all comes down to marketing ultimately because this is 2025, not 2005. Game engines are abundant and every game engine with decent amount of features will have a learning curve that typical game devs are always gonna cry over and complain no matter how gentle the learning curve is
Its naive of you to expect people to adapt a new game engine from a nobody dev even if its as good as unreal engine. They don't. The nobody dev's engine needs to be heavily marketed for greater reach among the target audience but even then, amateur game devs((like those /agdg/ fags who make small change) will not adapt it saying they are comfortable with unity or unreal or godot primarily because they already spent a lot of time and resources learning that particular engine and don't have the actual need to switch to another.
Just because a game engine wasn't well adapted doesn't mean that it can't make games. It might to the eyes of an inexperienced amateur who can't do shit without their game engine of choice and has no idea how their game engine works under the hood but things are not as simple as that. This kind of evaluation is too oversimplified.
Most of /gedg/ build frameworks for their custom game ideas(like OP picrel), not a full fledged engine and if their framework is capable of publishing one game then its well capable of publishing 1000 more variants. But publishing a game is of least concern when building stuff, especially for learning, not like you'd get it.

>>106132962
You're flat like paper mario but we still track cardinality (N,S,E,W) on the isometric grid.
- Single key down will move only on x or y axis: W=Up, A=Left, S=Down, D=Right
- Multiple keys down will move you on the isometric tile grid
- Caps lock will toggle a hold for the isometric movement (eg just W will move cardinal North on the isometric grid)
- Holding shift will temporarily hold the facing direction in place and move you on the isometric grid

>>106132962
>>106087598

>>106130875
Yes. Inspired by Transport Tycoon.

What kind of engine & framework do you need to make a Mario Party game?

>>106120386 (OP)
How much are you faggots getting paid to make these games?

>>106133512
about the same as you'd get paid for repairing your own car.

How's market like for graphics researchers, (not engineers)? Does anyone even want people like that?

>>106133893
small

>>106133504
Unreal Engine with heavy use of Nanite

>looking at minecraft source
>everything is some abstract abstraction
>looking for 1 hour and can't find any actual code related to the thing I'm looking for just delegations and abstractions
is this a good thing or just the work of pajeets?

>>106133504
stick to cocos if you are an absolute beginner.

>>106134662
that's what real-world projects look like

The book “Data Structures and Algorithms For Game Developers” by Allen Sherrod seems to be exactly what I’m looking for. A little worried about its old age but glancing at the table of contents there seems to be a lot of good stuff.

>>106134662
Any sufficiently mature codebase will become loosely coupled and abstract.

>>106134740
these are the only books we'll ever need.

>>106135112
the Eric Lengyel ones on top. Ignore the bottom section. I was too lazy to crop the pic.

>>106135112
Unless 2 has data structures and algorithms I’m not interested

>>106135149
what data structures and algorithms are you expecting to learn for rendering something?
You just need to learn math about visualizing and applying it. That is what the second book is mostly about, with sample codes.
Lucky for you, these code samples come along with data structures and algorithms in C++. But its the math you should be focusing on.
You will hardly find data structures and algorithms info on game engine dev tutorials because you are expected to have the knowledge of when to apply what solution to what problem appropriately and are also highly encouraged to always profile your solutions because in the real world, the stupid hardware drivers will always give different performance for same implementation.

>>106135233
I’m looking for renderer architecture not shader math

>>106135240
define renderer architecture
You just implement a tree that is compatible with glTF and USD file formats and forget about it, unless you have another definition for it.

>>106135252
Right now I’m just using regular vectors of shared_ptrs. What are the better container options?

>>106133905
I know supply is almost non-existent owing to the fact that everyone now uses Unreal, but are the people supplying getting paid?
Don't places like Kojima, Sucker Punch, Peal Abyss need them? NVIDIA has become full AITard, and I am not really interested in offline rendering, where the most demand is.

>>106135265
>What are the better container options
why?
if you are the only one that will be using it then just stick to what you are already using. Will keep things much simpler and honestly, I don't think what you are doing currently is bad at all.
Unless you want to have more compatibility with standard 3D file formats, going through gltf and openUSD docs and figuring out a common end point and implementing(an N-ary tree usually, which can be implemented with vectors) it for your use case is a significant amount of work. Its more fun to pick up the second book and implement different rendering techniques instead imo.
>>106135279
sadge

>>106135279
video game companies don't need graphics researchers, they exist but they don't work at video game companies

>>106135327
Because right now my application works on importing a single gltf at startup and not changing the items at all. But for a proper application I need to deal with stuff like deletion and insertion.

>>106135342
>video game companies don't need graphics researchers, they exist but they don't work at video game companies
Very uneducated midwit take. EA and Activision are pioneers in real-time graphics research. Ever heard of Pete Shirley.

>>106135417
I've seen some people in siggraph from game companies but I figured they were in academia

>>106135412
welp, gotta figure out a way to rely on whatever gltf library you are using for adding and removing more gltf scenes or nodes(however you prefer to abstract your meshes) on the go dynamically and chill, I guess.

>>106135417
>Ever heard of Pete Shirley.
Literally no one would hire him nowadays. People just want UE5 codemonkeys.

Can persistent homology be used in graphics?

>>106135075
this tbqh
procedural structslop only works on relatively small projects

>>106135846
literally half the world runs on what you call procedural structslop
Quake/Doom was made in C.
You can perfectly emulate dispatch in C.
The fucking Linux kernel is written in C.

Writing proper structs is modeling your data.
Learn proper engineering instead of thinking that the next abstraction will solve your shit without you thinking.

>>106135075
>Any sufficiently mature codebase will be full of hidden coupling and OOP spagetti.
FTFY
Plus a bunch of other bad (anti-) patterns due to organic growth.

>>106135846
>procedural structslop only works on relatively small projects
FYI unless you are using some esoteric or functional(pure or not) language, you are totally dealing with the "procedural structslop".
Java is a actually a procedural language with a retarded definition of OOP along with a brain rot syntax and semantics combo that restricts you from freely writing code outside of a class.
An absolute curse of an existence from every angle.

>>106120386 (OP)
>makes his engine
>doesn't make a game

>>106136756
>doesn't make a game
and?
the point of /gedg/ is to make an engine though
go back to /agdg/

>>106133148
there is no point in such evaluations without evaluations from end users - game developers
because a game engine is still a tool, not an end product
it might be high end feature rich super fast sota, but if it's impossible to use it's a bad broken nonexistent engine
making even the simplest game engine that you've used to make the game as a scripter, a designer, an artist, etc makes you better game engine developer than writing sophisticated pile of techniques never tested in real conditions
so no, there is no mocking
want to show me how good you are at game engines making - show me a game made with one of them for start

where's frogman?

>>106137066
>want to show me how good you are at game engines making - show me a game made with one of them for start
go hire a bunch of devs with your own money and request them to do that for you lmao
You ain't paying shit and even if you did, in case /gedg/ made an awesome game engine, your pay would be worth shit.
Bold of you to compare a product made by a self-sacrificing, passionate, solo dev to a well funded, corporate product.
Game engines are very complex and the most /gedg/ freebies can do is what they are doing already, building their own shit.
Making a game is dead cheap now with AI but even that same AI can't do shit when it comes to making a game engine.
As I mentioned earlier, you wouldn't get it.

>>106137093
me wondering the same.

>>106134662
grep is your friend

>>106136302
>Learn proper engineering instead of thinking that the next abstraction will solve your shit without you thinking.
The curse it the IT world. The next hyped pattern/tech/wtv will save us from technical debt and pajeet-level code.

>>106131098
>an AI can make a game but it can't do shit when it comes to making a game engine.
It's this the opposite?
engines are very streamlined and generalized, some kiddie was spamming a research paper github example with LLM trash making a physics simulation with vulkan and I'm personally surprised that the kid was able to get it compiling and they did all that in a week (i'm not actually impressed however, it's not like they are doing anything fancy with rendering / ECS, but it is a engine in a sense).
https://github.com/savant117/avbd-demo2d/issues/5
That being said, I think this just shows that Engine dev and game dev is basically the same in terms of LLM complexity. LLM games are just toys, LLM engines (games made without an engine) are also just toys. But there is a chance that could change once one kid makes $1 mill from selling a LLM game (if that hasn't already happened).

>>106136302
>Quake/Doom was made in C.
and they're simple games with barely any business logic

>>106136756
The engine is more interesting, I'm happy in my debug cube and imagining the rest

>>106141313
>simple
Not when I read the source.

So, is JBlow giving access to Jai now? Been seeing lots of people getting access. I don't want to write an email because I'm scared he'll call me a faggot or something

>>106142655
no it got leaked

>>106139402
>but it is a engine in a sense
a rigid body solver is in no way a game engine saar. Its just a mere physics simulator.
The repo you shared is just a bunch blocks simulated with physics for constraints like rope, rod and spring.
It doesn't even render sphere or anything else other than a bunch of cube.
And those cubes can't be controlled by other cubes at all. Physical collisions are their only interaction because its just a simulation. Stuff like one cube being able to generate hit events on another cube and that another cube responding to it is what it takes for this bland simulation to become a game engine.
I saw about that paper on 2 minute video, very interesting but assuming that this mere simulation would sell 1 mil would be too much imo. It just can't. It takes a lot more than this to be able to make 1 mil in 2025 as a game made through a game engine with this. Houdini and blender can do a lot of fancy rigid body dynamics than this.
>I think this just shows that Engine dev and game dev is basically the same in terms of LLM complexity
no?
Just because LLM takes same amount of compute resources to generate a response to any question of varying complexity doesn't mean the complexity of all the problems it gets thrown is always the same.
complexity of a problem is entirely subjective w.r.t the user giving the prompts imo.
LLMs are used to solve problems so a pro and an amateur can find the same problem to be simple and complex in their own perspective.
What makes game engines objectively more complex problem than and a video game itself is that a game engine is very low level while a video game is just a very higher level abstraction that relies on the game engine to interact with the underlying hardware.
>>106141803
skill issue

https://www.youtube.com/watch?v=RY0WclQU7x4
Extremely based, what's your excuse for not using Jai?

>>106142862
Why should I use Jai?

>>106142862
>Extremely based, what's your excuse for not using Jai?
Like I said, I don't want to email him

>>106142763
I'll see what I can find

>>106142902
Metaprogramming.
>>106142655
>>106142916
You could just use the leaked compiler.

>>106143036
>Metaprogramming.
Why do I need this

>>106142862
>what's your excuse for not using Jai?
why you asking this when no ones has access to it?
a leak of 6 month old version still means that no one has access to it.
keep on coping with your Jai shills bruh.
They can never get as better as my Odin shills

>>106143050
To avoid writing boilerplate code, having to maintain anything in multiple places (write once, generate all consequences), etc.

Granted if you're working on something small the benefits are marginal, but it's insane how much time it saves you on bigger projects.

>>106143116
>To avoid writing boilerplate code
but that's what llms are for

>>106143149
LLM boilerplate code runs at runtime.
Jai's metaprogramming code runs a compile time.

>>106143222
think really hard about what you just posted

>>106143240
>think really hard
sorry, I'm too retarded for that.
Can you spoil me?

>>106085981

loop cut, only segfaults if it doesn't loop back in itself.
camera can zoom and reposition to match the selection bounding box
matcap shader in the yellow ball is just wrong... i have some ideas on how to fix, but they involve trig functions which are suposedly expensive which defeats the purpose of using a matcap

>>106137393
> and the most /gedg/ freebies can do is what they are doing already, building their own shit.
exactly
shit, not a game engine

> Making a game is dead cheap now with AI
make a game then

>>106143572
>shit, not a game engine
yep. As I was saying, you wouldn't get it.
This is conclusion you can come to with your level of comprehension. Even grasping how complex a game engine could be is way out of your league lmao
>make a game then
I will when I feel like it, not when you feel like it.
For all that I care, making a game is boring. Making the shit that goes into making games is much more interesting.

Gonna try my hand at making a 2D DQ-like with a gameplay twist. Gonna prototype the twist first. Should I just use Unity (thus allowing for ez moddability for the final ver as well as faster iteration), or yolo it and do C/C++? I'm comfortable with both Unitydev and enginedev btw, I've nodevved with both options

>>106143638
no, it's you don't get it
you think you know what a game engine is, but the truth is you don't

>>106143792
okay

>>106142840
I think you are tunnel visioning too hard into engine dev must be low level and difficult.
You could find engine problems that are pre-solved, and you could find game problems that are pre-solved (AKA meaning the LLM should have many examples to reference from). But you could also find very hard unsolved engine problems, and very hard unsolved game problem.
I think arguably there are more unsolved game problems than there is unsolved engine problems, but I think because LLM's prefer projects that are small in code size, the engine / game separation doesn't really exist.
So in my opinion, every game made with an LLM that isn't for unity / godot / whatever, those are technically game engines. Sure three.js and JS isn't as hard as C++ and vulkan, but I think this is more of a user skill issue than a LLM skill issue (because it's expected for the user to fix the errors / bugs).
I think right now if you wanted to make something in a week, LLM's would be the way to do it.

>>106143983
>you are tunnel visioning too hard into engine dev must be low level and difficult
no, I am not. That is how it is supposed to be. Game engines used to be the entire operating system for a custom hardware
Just because our general PC are fast and game engine development practice using higher level languages like Python and C# became the norm, doesn't mean its ideal. As Mr. Jon Blow says, anything that isn't a manually memory managed, systems programming language is bad for video games
>there are more unsolved game problems than there is unsolved engine problems
There is no such thing as "game problems"
Its just implementing an ideas using an existing game engine so that makes the "game problems" you are talking about is the giame engine problems themselves. It is precisely so because whatever the "game problem" is, it requires distinct implementation in each game engine and if the features required for your "game problem" do not exist in a game engine then your "game problem" needs another engine that has those features. Even if you built your own framework that is tailored for your game in particular and isn't well generalized as a game engine, your "game problem" is still your game framework's problem. Factorio, songs of syx and terraria are good examples of game fraemwork's problems while no man's sky is a good example for engine's problem. Ideas are dime a dozen, its the execution that matters so its always the engine's problem with the game's design
>wanted to make something in a week, LLM's would be the way
yes. That is what /agdg/ fags should be doing. LLMs are perfect for them as making games is relatively dead simple compared to making engines
BTW >>106143572 is what an /agdg/ fag that pushes the ideology of finishing a game ASAP looks like
But when you want to learn anything fairly complex, you should do it at your own pace and once you venture deep enough, LLMs won't and can't be of much aid. This is what /gedg/ fags have to deals with casually

See ad for 970m oho better than 1050m I have so I go and ask for it...end up buying a 965m its the same card ugh but Its flashy and they're both 17"gaming laptops for vidya dev.

Books on vidya dev pls? To celebrate tinkerbell =) The ROG laptop I bought.

>>106136302
Most of the world runs on C++ actually.

>>106144748
could you speak in english

there is not a singe flaw in cpp

>>106144783

share books on video game development sir or mam

>>106144748
Personal suggestion >>106135112
ignore the bottom section. Eric Lengyel's book are very good.
>>106144782
not if you consider embedded shit
>>106144803
they why was it the reason behind James Gosling creating Java, GingerBill creating Odin and Jon Blow creating Jai?

>>106144812
if you are looking for a books on doing game dev using game engines like unity or unreal then you are better off following tutorials on youtube and reading their docs.
The books I from my previous post are for game dev in general.

>>106144827

no bundled for starting how to programmings?

>>106144861
ah, you are an absolute beginner.
Forget about the books from the previous post.
Just stick to using the raylib library. There are plenty of tutorials on youtube to guide you through.
sample codes - https://www.raylib.com/examples.html

>>106143657
honestly if you really want *le ebic fast iterations* use love2d.
unity is fucking garbage its genuinely faster to develop a game in c than in unity.

bump

>jblow thinks his game is "meaningful" and will go down in history
>jblow compares himself to the greatest artists and scientists who got respected after their death, implying his little block game will blow up in 70 years
>jblow is a christ cuck
i cant
https://www.youtube.com/watch?v=RY0WclQU7x4

>>106144543
> BTW >>106143572 is what an /agdg/ fag that pushes the ideology of finishing a game ASAP looks like
> But when you want to learn anything fairly complex
if you want to learn anything fairly complex you need to learn in steps and have a foundation
learning game engines without making them and testing is like learning programming by just reading books without writing code
i am not an agdg fag, i'm telling you this from the perspective of a person who works with both engines and games

>>106120386 (OP)
What's your cuckiest gamedev question?

>>106147042
>learning game engines without making them and testing is like learning programming by just reading books without writing code
the comparison is not quite accurate but nobody disagrees with that.
This step by step learning could start from math and progress along with programming, computer graphics and graphics APIs, real time rendering, game object system and finally towards a game engine and refactoring it there after.
Or the step by step learning could alternatively start from directly building games with an existing engine, learning a graphics API and math simultaneously, achieve doing real time rendering with no clarity but a general idea on how things work, iterate the last two steps again and again till you are comfortable and then slap a game object system like actor-model or ECS and make a game engine.
Learning happens in progression and it doesn't matter on the long run how one starts and proceeds, as long as they are consistent.
>i'm telling you this from the perspective of a person who works with both engines and games
then why are you saying games and game engines are the same thing in terms of complexity?
They are not, both in general and even for LLMs.
If that was really true, /gedg/ would have equally or more engines made than the number of games /agdg/ fags are making.
The only reason why there aren't many /gedg/ engines is primarily due to the complexity caused by graphics APIs and manual game asset resource management shit. A game dev using a game engines doesn't have to deal with any of that and just chill with a one liner code.
No matter what complex of a "game problem" you are thinking of, in the end it all depends on how well a game engine could solve that "game problem" for you. Choosing an ECS game engine is better for some "game problems" and OOP engine for another set of "game problems". Choosing the wrong engine will only make implementing your "game problem" much harder than it needs to be.

>>106147243
> This step by step learning could start
> Or the step by step learning could alternatively start
> Learning happens in progression
but there must be practice
making a game is the practice
not a full game, but something more advanced than a tech demo, where you or someone else test it as a user
because again it's a tool and usability is a crucial part of it

> then why are you saying games and game engines are the same thing in terms of complexity?
i never said that
and i don't care about your and agdg relationships

>>106144803
Module support isn’t there yet otherwise I agree. Also reflection coming in c++26 is huge.

>>106142862
>>106146970
>>106147056
>Christian
Uh oh, this is gonna send the kind of person who hates Blow into an absolute rage meltdown

Is there a better container than std::vector for sequential writing and accessing?

>>106147531
Maybe std::dequeue. Otherwise vector is simply the best.

>>106147528
christians dont have a monopoly on god

>>106147632
He doesn't strike me as muslim or jewish.

>>106147692
hes just a theist or a gnostic

>>106147743
Those schizos are an absolutely tiny fraction of those that believe in capital G God. Unless you have any proof, why would you assume he'd be one of those rather than just Christian?

>>106147764
if you watch his videos you can get a good idea of his beliefs

>>106147485
>but there must be practice
exactly what I meant by being consistent
>i never said that
okay. whatever >>106143983 was on about
>>106147531
std::vector is the best you got in C++.
If you don't like it, a more flexible solution is doing things C way by creating your own struct and maintain it(like allocations, deletion and freeing operation) manually.
Kinda like how Frosch did it in om
https://codeberg.org/FroggyGreen/om/src/branch/master/engine/src/ecs.h#L9
>>106147632
all monotheistic religions have a monopoly on god saar
Especially them abrahamic ones, each of them has their own exclusive gaud.

>>106148643
you can believe in god without being a member of a religion

>>106148717
99% of the time that's just
>I've realized God exists but I don't want to make any major changes in my lifestyle so I'll come up with my own version that conforms to my opinions and habits
Alternatively
>I'll come up with my own version that lets me have full control over its tenets and get money from people I con
if you're a pastor in flyover USA.

>>106148911
why would god require you to change your lifestyle? you know that humans invented religion right?

>>106148928
Case in point.

>>106148963
???

Morning bump, I am blessed to be in this realm for another day by my Lord Jesus Christ

Gotta implement a byte buffer for my udp network component and all the in-place memory management options suck for units that small.

Working on my stripped down simple ECS.

Surprisingly more work than expected.

>>106146970
>christian
clearly, he is talking about vinshu

>>106152898
LoC?

I finally implemented double precision, but the code looks ugly now because the graphics part must use floats while the game logic must use doubles. I feel like going back to floats now, it made things more seamless. I guess it's nice that I have a safety net now.

>>106137393
>me wondering the same.
we know it's you.

>>106154472
500 or so, but will probably be 1-2k total. Not having archetype migration helps a lot.

>>106154498
>put anchor position in world
>when rendering something make the model matrix relative to the anchor position
what's the problem

>>106155111
I need some objects to exist in specific coordinates very far away from each other. Big coordinates are unavoidable. Anyway, it was a failure. I'm going to have to put a cheap precision test on my objects to warn whenever precision is significantly compromised because the implementation caused a weird bug that I don't feel like fixing. Despite that, I'm relieved I won't have to deal with doubles anymore.

Do we really need to waste space storing size/scale for 3D models and matrices? i guess it's useful for particles but when do you actually scale normal models or even skeletal bones?
scaled up/down models look like shit, they either become too detailed for the rest of the scene when they're small, or look like shit with pixelated textures when they're big.

>>106157416
Not sure what you mean "waste space". It's baked into the transformation matrix, there's no extra space that it takes up compared to a rotation-translation matrix

my data structures involve a bunch of pointers. now i'm realizing this was a mistake, since i'll have to write them to a file...
sure i could convert them to some sort of numerical index before saving, but if i'm going through this trouble to save, then i should probably just have them in memory as arrays and indices

>>106157776
That's what I do, and I wrote my own code generator to create those structures, so I can Adjust it to save to files too.

With my scene nodes being a vector of shared pointers I realized that instead of sorting render types in the draw function I can just make a vector of weak pointers for each render type and sort before drawing. The real question is whether I can do it in “immediate mode” which means sorting in between every single frame instead of only updating when scene hierarchy or materials change.

>>106150049
what is that phone with a keyboard like that?
I don't mind believing in Jesus Christ if he blesses me with phone.
>>106154579
In your dreams.
I do work long hours but not as a security guy and besides, why would Frosch choose to role play a pajeet saar?
Its probably more like he's chillin' 'cause I'm here to gatekeep /gedg/ from AI shill titles for the time being.
>>106158865
>can do it in “immediate mode”
stuff like sorting is fine if its just a handful of elements.
If you think that it is something that doesn't need to computed every frame, then hide it behind a boolean flag that resets itself by the end of the frame. This ensures that the expensive operation only last through out the span of the next frame only after getting triggered in the current frame.

bump

>>106157776
Is it possible to write down an entire struct as text? I want to implement engine-wide save states.

>>106161912
In hex

>>106161912
>engine-wide save states
what states are you trying to save?
You just figure part that out and decide on what file format is the most compatible for you when you write all these save states into a file(or files).
Its fine totally to rely of sqlite for this as well.

>>106161912
Saving as binary is easy. The hard parts would be handling pointers, gpu resources and I have no idea what would saving 3rd party classes look like.

>>106161912
https://msgpack.org/index.html

bump

>>106162737
>strings require only one extra byte in addition to the strings themselves.
where have I heard this before?

Now I can route track, now working on actually placing it.

>>106165736
\0

>>106159806
Its an fx tec pro 1x running droidian

>>106166159
What are you using to make this?

>>106120386 (OP)
gonna be solo dev on a project with two artists, mecha shootemup top down. should i go with godot cause it has tools for non technical artists to fuck around with and see how the game feels? I have exp with BevyECS but im not sure thats the right choice cause rust is kind of ass for games.

Any thoughts on using SDL vs Raylib? I've used SDL a lot before and it's a very good library, but Raylib seems pretty nice too. But I've heard for real and bigger projects that Raylib has a lot of issues that you wouldn't have to deal with using SDL.

>>106143305

finally matcaped succesfuly
the secret is to put the viewspace normals in "eyespace", that is, instead of making them local to the camera, you make them local to the ray from the camera to the fragment.

Does audio programming depend on the operating system?

>>106169203
I dont think so

>>106169203
everything eventually depends on the OS

>>106167393
Your chances of finishing and publishing the games will be higher when using Godot.
When using Rust, forget about completing, your changes of abandoning the game itself will be much higher. That has always been the fate of almost all the Rust projects that were fairly complex but weren't well funded. Eventually, nobody would feel like dealing with unnecessary complexity for getting shit done shit, unless paid.
>>106168688
>Raylib has a lot of issues that you wouldn't have to deal with using SDL
makes sense. Raylib is a simple and easy to use library and SDL on the other hand is a bare-bone low level library. Its very comfortable to get started with Raylib but in terms of stability and performance, SDL beats Raylib as it offers more flexibility but SDL also requires a lot of work for achieving that.
>>106169203
yes. Just use OpenAL and chill
https://www.openal.org/
>>106168953
peak

>>106161912
Compile it to an array of integers.

>>106167393
yeah, go with godot unless you wanna build the tools that allow your artists to fuck around... you will still write some tooling, but those are engine extensions 50 to 200 LOCs usually very easy python-like gdscript
you should really look up the features that you need when picking an engine

>>106167387
Transport Tycoon style railroad game.

>>106122516
which AI sub is best for engine dev?

>>106128509
I have seen this years ago, I think where you added the cloth and physics? how fucking long have you been working on this demo?

>>106172051
thats a generic 3D scene used by lots of people

>>106172056
oh lmao, then that makes sense. Where is this scene from exactly?

>>106169553
still coping about ai?

>>106160259
google's automatic game generator just previewed. surpasses grok 4

>>106172069
fuck off retard

>>106172069
> google's automatic game generator just previewed. surpasses grok 4
but can it do pixelart

>>106128509
API?

>>106172051
>>106172059
its the sponza model. an old crytec tech demo scene

>>106172145
mainly directx 12 and win32

Been using ludas in ue5 for a bit. It helped fix some bad timers i set up and exchanged for event dispatchers instead. I had a basic ai system set up and it helped me build a threshold for following and attacking the player. I have a long way to go but here is an example

It never occurred to me that I could create a buffer larger than the data at hand which fixes my whole payload issue where the size of data can change between any frame. Now I can casually resize my payload vector without having to worry about buffer recreation. Eventually I will have to add size checks to recreate it if necessary but for now it just works. I can add and remove lights effortlessly.

>>106172071
go ahead keep seething about ai, until some teenager completely destroys your engine you spent years on with a ai generated masterpiece in under a month

>>106172602
whens that gonna happen

>>106172069
>still coping about ai?
Gatekeeping the tread is already handful. Got no time for that shit.
You better lay low unless you want to get cooked.
>>106172165
noice.
Its giving Hood: Outlaws & Legends vibes.
>>106172296
arena allocator to the rescue, huh

Hey tourist here, do you guys know of a poster here who streams gamedev and makes autistic posts and possibly avatarfags?

>>106172960
Thanks anon, I wish I could make a game like that lol. Lots to do but I will keep at it!

>>106168688
>SDL vs Raylib
That's like asking about C vs Python.
>But I've heard for real and bigger projects that Raylib has a lot of issues that you wouldn't have to deal with using SDL.
Try rendering a mesh on your screen with Raylib and you'll see the problem. It's so extremely high level you have zero control.

A better question would be Raylib vs Sokol vs bgfx, but here too Raylib is so high level it's not really comparable.

Spent a buncha time finally fixing the opengl renderer for the windows build of my engine. I'm using Avalonia and it uses ANGLE behind the scenes which is awful for any moderately heavy game engine stuff and the native windows opengl backend for Avalonia is buggy and crashes like crazy. Thankfully Avalonia has a way to use gpu interop to render to a Direct3D texture that you can display, in this case I had to use NV_texture_interop to render to a framebuffer in an opengl context in a hidden window that is mapped to a direct3d texture. Before the animations would jitter and skip frames because the avalonia opengl context was buggy as fuck and some textures would just fail to load.

>>106172165
Updated npc armor set, has different dialog as you go through his prompts and repeats last line. Also spawns an item in my character's inventory.

>>106174104
why does this look so bizarre to me. rotoscoped? tweening? ai?

what do we think of sudo loveme baby ?

>>106174104
>clown
>jiggling honkers
would.
>>106174398
>Also spawns an item in my character's inventory.
damn. Seems like ludas isn't as bad as everyone's making it to be.
>>106174457
>ai?
if that's the case, artists are really cooked.
Her eyes when the player is leaving were full of soul and very alive.

>>106174565
Works well for me, I try to figure it out before asking the agent but it helped me build a competent save and load function.

i just put trigonometric functions in the fragment shader and it's not so expensive desu

>>106174693
ah, I see.
AI is well capable of generating code for a game engine.
But since AI wasn't this good a while back, ludus sucking when it first came out checks out.
>>106174929
I put bezier cuves, lots of them for font rendering, in fragment shader. The performance was not that bad but if I were to compare the frame times of
(no computation on CPU with bezier cuvers in fragment shader) vs (delaunay triangulation on CPU along with GPU renderer present time) for font rendering, the former was less complex and slightly faster most of the time.
But putting bezier curves in fragment shader in general is a really bad idea. It just worked out somehow for my use case.
So ultimately, it depends.
Using trigonometric functions in fragment is not bad in general at all. Just look out of for branching. Its better if the any branching present in shaders is as uniform as possible.

>>106175129
Oh yeah initially, but now the conversation limits are much longer than they were and it gives me more options than it used to but yeah it did suck at first.

bump

feeling the urge to dust off my abandoned win32+dx12 c++ project

Say I have a bunch of concave polygons in 2D space. Some might overlap or be fully contained in another. Specifically states/cities/regions on globe and GPS coordinates.
For a given point and I want to efficiently find all polygons that contain this point.
I started with simple point-in-polygon check (actually cute and elegant algorithm) with for-loop over all polygons. Just pre-calculating AABB for of each polygon and pre-checking that yielded over 50x speedup.
But I want to scale it up - more detailed polygons and way more of them.

What data structures/algorithms could I use? I obviously found binary space partitioning trees, quad trees, k-d trees, ... but everything I found only inserts points into them and queries to find near neighbors. Not shapes and doing hit check. Any suggestions?
Obviously I could again use AABB for those trees and do expensive point-in-polygon on just few candidates.

>>106177240
Maybe RTree? I don't think there should even exist a point variant so it should be easy to find the proper implementation where you insert rectangles. For static zones it outperforms quad trees.

>>106177452
quick googling actually found few geospetial applications so it's probably the right choice, thanks Anon

.

>>106176715
i think thats a good idea

man, making a custom engine for your game is quite manageable.
making an engine to release as a standalone product? fucking brutal. website and docs are just as much work as coding itself.

>>106181517
After some time you will need the docs yourself

>>106135112
Where can I find all 8 books? I could only find the first two.

I was debugging a memory leak yesterday for spell animations. I confirmed the logic to delete entity was good so I was stumped for a moment. Eventually I realized I wasn't passing in a uid I generated for the spell entity just before I created the spell, and the default of no arg passed to the class is for it to generate a random uid for itself. With the mismatch the callback in the animation is never properly culled. Fixed now.

>>106181517
yeah, it's just a glue to different libraries, an as generic as possible ECS and a default gameloop
if you wanna get fancy, throw in a GUI and tileset editor. 3D editor could be third partied to blender + geometry nodes

>>106185137
>blender + geometry nodes
its better to completely avoid touching geo nodes imo
Just do create a blender plugin where MESH and skeletal data of 3D objects in blender's scene are sent trough a pub-sub mechanism, like the typical blender live-link plugins for unreal and plasticity.
This way, the engine just has to only fetch the 3d geometry data and present it. All of blender's bloat stays inside of blender with no hassle of export-import bs. Its a win-win.

whats everyone working on right now?

>>106186821
Im working on my retarded enemy ai.