[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


File: 1766538660639238.png (24 KB, 1556x1556)
24 KB
24 KB PNG
it's over 9000 extensions 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.

Previous: >>108436047
>>
first for procrastinating out of fear for the complexity of the next feature you need to code
>>
Second for massively overcomplicating your life by following weird ideas you found on Digital Grove which sounded great in theory but turned out much harder in practice.
>>
trying to recreate the shader from youtube vid rel entirely as a post processing inside Unity 6 SRP.
on the surface shader, its not a regular shader, and instead outputs 3 channels for data. which is also why the guy is pink at the start, but the rest are black, I still need to improve the channel information that is sent from here to the post process.
this lets me use a gradient texture inside the post process, to determine how is he colored and how to use more than one color for my entities
then there's a separate pass to project noisy textures, and is how the guy has stable noise on his body, whereas the rest of the scene uses easier to set up triplanar textures.

here is the video i am trying to base my style after:
https://www.youtube.com/watch?v=xmS5aQEA1Xw

I think overall, as a post process, it's cool. I think the effect would be a little easier if I set it up as regular surface materials with _GLOBAL properties and control from a script or something.
Also, I dont know how he did that 16-directions rotations for the one guy who has quantized rotations in the video. Tried my best, but I'm no good at quaternions.

do you like this? would you play a game like this? My goal with this style is to recreate the old Nintendo Power Legend of Zelda art paintings by Katsuya Terada, combined with comic books style like HellBoy
>>
File: cross-over.jpg (85 KB, 488x1418)
85 KB
85 KB JPG
Finally got somewhat decent detection of overpass roads and able to color the underpass road and lane dashes differently than the rest of the path.
Testing with 3 varying elevations with road rendered on a single path.
Am I insane to try to render everything?
Because I like kt better and it is making the full game dev way easier down the road.
>>108473375
This looks cool. The guy even has /g/ forward head posture.
>>
People who are doing things in DX12/Vulkan, did you start out with higher level APIs?
>>
>>108473259
Ah, the ol' one-two Blow-Muratori knockout combo
>>108473772
No, I dove straight in. I didn't make my own engine to rely on black box APIs, so I didn't want to use pre-12 DX or OpenGL. Having to deal with the driver being a black box is already annoying enough, don't need another layer.
>>
>>108473772
yeah
>>
If you could go back to college what major would you pick if you're into game dev?
>>
File: 1758767662189980.jpg (219 KB, 768x768)
219 KB
219 KB JPG
>>108474163
CS major then take some music, writing and drawing and 3D animations extra private classes.
>>
>>108474163
There is no if, I'm in the middle of doing an online compsci degree at age 32.
>>
>>108474163
Nothing, you will learn more by yourself and it's free
>>
>>108474163
fuck that would be a waste of time and money
>>
>>108474163
You don’t need an education for video game development. Just lots of patience.
Brush up on your math skills.
>>
>>108473772
i had previous experience with dx 10 and 11 before i jumped into dx 12
>>
>>108474163
nothing, i have no talent for art and the cs course was a fucking joke, practically didnt read anything at all and still got great marks
>>
>>108474163
A gamedev major. My country has a few decent ones actually.
>waste of time
Technically yes, but you're doing it for the line on your curriculum.

Though the biggest would probably be just starting with programming 10-15 years earlier.
>>
>>108472901
What extension does Valve, Qcom and Android have for gamedev?
>>
>>108474288
what
>>
File: 1741969849027067.gif (105 KB, 500x281)
105 KB
105 KB GIF
>>108474288
I was just going to post a link to the Vulkan registry, but for some stupid reason, they removed the extension specification list off of there.
Well, I guess you have to look in the actual docs now: https://docs.vulkan.org/spec/latest/appendices/extensions.html
>>
>>108474163
if I went back to college I'd do the same thing (take CS) but i'd cling onto my Graphics Programming professor and pester him for industry connections instead of being an asocial incel freak who only cares about passing assessments and exams
>>
>>108474380
Industry connections for making indie games?
>>
>>108474393
industry connections for graphics programming gigs, like engine development
>>
>>108474346
>read these whitepapers
fuck off
>>
>>108474415
Why would you want that when you're making your own game?
>>
>>108474346
https://vulkan.gpuinfo.org/listextensions.php?age=historic&apiversion=all
>>
>>108474420
are you retarded? hes talking about how things could have played out, getting an industry gig right out of uni. not to mention lots of people who works in the industry have hobby projects
>>
File: 6dof_slopes_comp.webm (3.63 MB, 1024x768)
3.63 MB
3.63 MB WEBM
slopes working again
>>
File: kyle.jpg (21 KB, 1078x516)
21 KB
21 KB JPG
>>108473462
thanks, its a model called Unity ADAM. usually I'll use robot kyle (pic) or a spiderman model for working on shaders but I used this ADAM one cus the video I wanna recreate is using it too
>>108474163
I would not go at all. Costs too much $$$ and I still owe them $$$, and even with the $$ I spent, I cannot get a job that will pay enough to repay them.
>>
>>108474501
realtime rendering discord poster
>>
>>108474501
What's the trick behind 6dof in a raycaster?
Also how did you implement such good-looking slopes? Any resources?
>>
>>108473772
Nope. I figured I was going to learn Vulkan anyway and just made the deep dive. It was rough at first but very much worth it.
>>
>>108474517
its tricky as fuck, and honestly I still don't 100% understand it.
https://github.com/sp4cerat/RLE-based-Voxel-Raycasting/blob/master/Paper.Efficient%20High-Quality%20GPU-Based%20Visualization%20of%20Voxelized%20Surface%20Data.pdf

this should help give you an idea

the main trick is that you split up the world into 90 degree quadrants in world space, and map those to different offscreen buffers, which allows you to raycast to contiguous columns, for performance.
you then map those offscreen buffers onto the screenspace area covered by the world-space quadrants.

voxlap is very similar to this (although with voxels instead) but it's clearly different in the way it maps the quadrant buffers back onto the screen (and it uses a ton of hand optimized assembly, allowing it to run on a pentium 3..)

>>108474517
the trick that makes them very easy is texturing floors/ceilings in vertical columns rather than horizontal rows. you lose the constant-depth benefit, which means a division+mult per pixel, but modern cpus seem to handle that no problem.

you simply project the calculated start and end height of a slope for a given ray intersection, and fill texture between those.
>>
>>108474417
In reality, most extensions that aren't *-KHR aren't super relevant to most general use cases.
>>
I lost, I fucking lost guys what the fuck
>>
>>108474417
Sorry, Vulkan is whites only.
>>
>>108473772
Started out with OpenGL ES 2, then went to Vulkan 1.1 years later.
>>
>>108474622
Except nvidia owns 95% of the GPU market on PC. You'd be retarded to skip over their extensions, especially when most of their extensions get officially adopted eventually and you'd have 90% of the work done by implementing their extensions. Like the entire raytracing API for vulkan, mesh shader extensions, DGC extensions, etc. Even when they're not 1:1 you'll still have most of the world done
>>
>>108474663
Hey can you be more respectful?
>>
>>108474631
the game? me too.
>>
>>108474708
Fuck you.

I went 15 years. I just lost the game
>>
>>108474663
Using extensions is risky business.
Either you
- Literally lock yourself to a vendor, or even worse, a specific generation of cards: which is fucking retarded if you're making anything you intend someone else to use and you're literally not targeting a single console.
- Have to create multiple code paths with an exponential growth of configurations which may or may not be present, which massively increases code complexity and makes it far harder to test.
If you're a big studio with the resources and the need to actually truly push maximum performance, maybe the 2nd option is viable, but lets be real, you're not.
With the really big "fundamental" shit like ray tracing or mesh shaders, the answer is always to wait for the KHR version to come out.
>>
>>108473375
Pretty cool
>Also, I dont know how he did that 16-directions rotations for the one guy who has quantized rotations in the video.
Someone made a tool that does that in UE4, so I'd suggest you try finding that and see if you can rework it for unity.
>>
File: my engine.gif (600 KB, 1344x676)
600 KB
600 KB GIF
more progress, chuds.
>>
>>108474988
>Nim
Based Wirthian Chad
>>
File: 1696959565041831.jpg (60 KB, 1200x800)
60 KB
60 KB JPG
I just realize I can use a system similar like mugen where I store each animation inside something like anims/idle where each folder is a substate animation.

Then on the JSON of the player I just need a seq[string] to store the animations, while are just filepaths to the subfolders inside player.

Then I just on my player just hold a 2D array of textures.

And then I just loop then as an animated sprite in naylib.

And then add some basic reserved keywords like IDLE, WALK, WALK_F, WALK_FD, WALK_B.

Or maybe store all the directions of WALK inside a folder called WALK and then each direction is just WALK/F, WALK/B

That way adding newer states only requires to add newer folders and inside the .webp and a basic edit to the JSON.

It would be much much easier and simpler than all the autismo of godot and unity.
>>
File: 39.png (24 KB, 121x134)
24 KB
24 KB PNG
>>108475193
>.webp
>>
>>108475193
>Quest complete: hello world
Keep going Cris, you will get there
>>
>>108475228
webp is actually great because is like 0.1 of .png and is also lossless.
>>
File: 1604254866663.jpg (112 KB, 445x503)
112 KB
112 KB JPG
>>108475305
>>
>>108475193
>>108475305
>reddit spacing
>thinks webp is a good storage format for textures in a game
I can't even even
>>
haven't worked on my game for a month. I'm stun locked by art :(
>>
>>108475518
It's cris. Any time you see schizo babble with a cat picture attached it's probably him.
>>
>>108475534
just draw some placeholder things you can replace later
>>
>>108475534
I've been stun locked for a decade anon, part of the process
>>
>>108475518
what's wrong with webp?
>>
>>108475770
it's a format almost nothing supports that exists solely because of google shilling it hard
>>
>>108475790
raylib supports it.
>>
>>108447717
*yap:
Issues I found with llvm-mingw using pdb files:
-C++ types in libc++ are not visualized by debuggers, someone made a .natvis libc++ visualizer for chromium, so I just renamed std::__Cr with std::__1 and it worked.
-vcpkg wont copy the pdb files for me (locally it will still work because the pdb path is hardcoded in the DLL, it also saves disk space).
-LLVM 22.1.0 added in a feature for if you use ubsan to output formatted errors with traps enabled. Usually traps are vague and unhelpful, but with __builtin_verbose_trap it will store the message on the stack. This does not work with pdb files (including msvc clang, llvm installer).
Also the __builtin_verbose_trap will not work if you use -g or -gline-tables-only. Meanwhile a release build without debug info will still have functioning stack traces. (on msvc/msvc clang you need a stripped PDB...) But the debug trap is not a real function, so it shouldn't exist on the non-debug section (it always appears as inlined).
So... just re-implemented ubsan using the micro ubsan bsd code (so that I can use it for making reports, I might try minimal ubsan for production). And I will use msvc/msvc-clang because of cleaner stacktrace code, the binary size is much smaller, and iterator checks. (msvc is slower to build + no colors in my errors, but msvc-clang is not perfect with optimized debug info... BUT that's a nitpick).
Also it seems like llvm combines ubsan's runtime with asan's runtime, but that breaks my custom ubsan runtime with duplicate symbols... Asan already has a callback I can report with, but not ubsan.
>>108447096
>llvm-mingw defaults ALSR off (for addr2line?) so I add link: -Wl,--dynamicbase,--high-entropy-va,--nxcompat,--large-address-aware
I was wrong, ASLR is on, this isn't a recent change or anything. The other flags are already enabled.
Also libc++ on llvm-mingw is built with cfg guard, not particularly interesting (I use cfi-san) but that's not vcpkg's default. I used PESecurity to check.
>>
>>108475534
That's what 'programmer art' is for.
>>
>>108476262
What is it about build systems that cause people to write paragraphs and paragraphs of useless information
>>
>>108475534
https://perchance.org/free-text-to-image-unfiltered

It doesn't have to be good. it doesn't have to be permanent. It just has to be.

If you're after 3d models, then the default blender shapes, and mixamo models are your best friends.
>>
>>108476411
>>108475534
kenney assets are decent enough too. pretty sure there are 2d and 3d packs on his website.
>>
File: my engine.gif (639 KB, 804x623)
639 KB
639 KB GIF
I'm so happy now, chuddies.
>>
>>108473259
Which ones?
>>
How would someone who knows what they're doing in OpenGL make a way to blend two different textures in a smooth gradient, kind of like how there's a smooth mix of colors in the Hello Triangle?

Followup question, is it even worth it to bother? Is it just unnecessary fluff to a shader program that could be better done with less intrusive methods?
>>
>>108477308
sample 2 textures and use mix()
>>
>>108477335
That was my initial thought.
Right now I'm trying to troubleshoot an implementation where my vertex shader has texture coordinates, an index to texture 1, texture 2, and a float to be interpolated through the fragment shader. I'm using a 2D texture array to store an atlas of sprites to change between.

It's kind of a lot to send to every vertex of a mesh, in addition to stuff like position data and color change. Or maybe it's not that much, I don't know.
>>
>>108477308
>fluff to a shader program
> less intrusive methods
you have no idea how this works
>>
File: test.webm (3.02 MB, 1280x720)
3.02 MB
3.02 MB WEBM
Putting cube in world using CHICKEN.
Probably will need a few helper fucntions to shorten the time.

I've shown this before with the old engine. But now at this point I can say the new engine has caught up with the old engine.

Time to move on with new things.
>>
File: 2026-03-29_00:30:14.png (210 KB, 640x480)
210 KB
210 KB PNG
>>108477335
ok, done troubleshooting it. here's what it looks like when it's working well. Still not sure if I want something else for this.

>>108477354
if I did, I wouldn't have specifically asked for "someone who knows what they're doing in OpenGL"
I know that I don't want to send mountains of data to the GPU for each vertex every time I load up a VBO. And I don't want to go out of my way to create some convoluted way to achieve something that someone else already thought up an easier way to do years ago.

blah blah premature optimization, yadda yadda, getting it working is more important than getting it working well.
I know, but I also know that I want to learn it right.
>>
>>108477383
Shaders is the least intrusive way to do any graphical effect
You aren't "sending mountains of data" you're sending a single value per vertex
>>
>>108477391
do you mean in this context, or just in general? how is it a single value when you can have many different attributes in your buffers?
I'm not even trying to be snide about this, I'm actually asking. one of the first bits of advice I heard was to watch how much you're sending to the GPU.
>>
>>108477405
How are you determining the blend value? Either you do it algorithmically in the shader and you don't need a blend value, or you send one in
>>
>>108477419
I was already sending an int for the texture layer, and now I'm also sending another int for the second layer, and a float for the blend value.
It might not be a lot, but it'll certainly add up, and I want to at least be efficient where I can.
Before I had the idea to do this, I was trying to find an easy-ish way to not have to send that same layer int to every vertex when I could instead send it to each polygon drawn. Also considering if that was even worth the effort.
>>
>>108477444
There's no per-polygon attributes you only have per-vertex
I take it you're trying to make a multilayered terrain, I send an entire color (RGB) aswell as the blend value for every vertex
Someone told you " watch how much you're sending to the GPU" but they didn't give you any metrics so now you're trying to your absolute hardest to save memory like a dumbass
Do the math, look how much memory your GPU has, look how much memory a vertex takes and how big your terrain is, you can afford more, my regular mesh format uses 64 bytes a vertex
>>
>>108477457
y'know what, fair enough. It's not like I'm going to be making a graphical powerhouse or anything. It's even designed to mimic the style of old games from the 90s.

I'm not all that concerned about this specific thing taking up too much video memory, or tanking the transfer rates between the motherboard and GPU or anything. I was just wondering if there was a bit more of a popular and efficient method for doing this.
>>
indie gamedevs on modern platforms using a retro aesthetic are posers
real ogs dev for an actual retro console
>>
>>108477858
this is true
don't let them hear it; they will be happier in their blissful state
>>
>have a question
>consider asking in 4chan
>already expect a snarky unhelpful response or waiting hours for no answer
>remember that AI exists
>get detailed answer immediately
>move on to the next thing
The future looks bright.
>>
It took some fiddling with Slang but I finally got vk_ext_descritptor_heap working in its entirety, along with shader objects. No more pipelines, pipeline layouts, descriptor sets. I no longer use the primitive (set = 0, binding = 0) that you midwits use. Everything in the shader is accessed via DescriptorHandle<T>.
>>
>>108478055
>contemplates attention seeking
>does it anyway
Do you want use to give you a cookie for finding out you have a brain
>>
>>108478164
Sorry you were replaced by AI.
>>
>>108478055
"you are absolutely right" - any ai ever
>>
>>108478055
You're just too stupid to use 4chan properly.
>have question e.g. "Should I use GLSL or HLSL?"
>ask normally
>*crickets*
>second try, make completely baseless, cantankerous claim like "GLSL is shit for GNUtards, use HLSL it's backed by a proper company (Microsoft)"
>300kg autistic neckbeard sperg in his mother's basement with 30 years of experience in the matter writes 5 max length posts explaining why I'm wrong
Easy.
>>
>>108478258
Reminds me of using a search engine and having to spend 2 hours crafting a perfectly roundabout query that will prevent you from landing into responses about pop music and car brands.
>>
>>108478258
to start with, glsl and hlsl are api dependent, unless you are doing some wierd spirv compilation. the second is just classic "im correct because im correct" trash that is so common in these threads
you shouldnt really ask for opinions on what languages and apis you should use, that shit always ends up like that. ask for reason to use something and advice on how to do something
>>
>>108478312
>glsl and hlsl are api dependent, unless you are doing some wierd spirv compilation
Both compile down to spirv, what's "weird" about it?
>>
What is the most complex mathematics you used for your projects?
>>
>>108478258
You should get a xilinix and learn verilog to make your own gpu to be able to use your own shader language.
>>
>>108478807
not quite yet, for hlsl they are waiting for sm 7.0 to support spirv, otherwise its dxil or something older
>>
why is the pow function so fucking slow

i'm making a CPU ray-tracer and using pow function to gamma-correct tanks the FPS
>>
>>108478258
>>108478312
kek
>>
>>108479746
A multiplication should be like 5 clock cycles, a lot less than the 3 billion cycles per second of most modern CPUs.

How can it be slow?
>>
>>108479782

i think there's some compiler issue i'm not understanding

on gcc it's SUPER slow
on msvc it's much faster than gcc but still slow
>>
>>108479843
Absolute worst case just do it in ASM, no compiler fuckery there.
>>
>>108479890
>>108479843
>>108479782
>>108479746
>A multiplication should be like 5 clock cycles, a lot less than the 3 billion cycles per second of most modern CPUs.
POW is possibly many multiplications and it's not a instruction-equivalent so it's up to the complier.
If you're doing low powers first thing to try is just doing the multiplications "manually". x*x instead of x^2.

>i'm making a CPU ray-tracer
Well... That is always going to be slow. We're just at the point that $2000 GPUs with dedicated hardware for it can do it at 30fps.
>>
>>108479950
>If you're doing low powers first thing to try is just doing the multiplications "manually". x*x instead of x^2.
Compiler already does that specifically for exponent of 2, or at least msvc does.
>>108479843
Try Godbolt and see the outputs. Is it even getting vectorized?
>>
>>108479746
pow might be using a log optimization instead of the normal exponentiation
>>
>>108479843

turns out pow is retard slow for values between 0 and 1 and exponents also between 0 an 1 and gcc is slower than msvc and clang at that

solution is to generate a table of pre-calculated values
>>
File: 1668464446668.jpg (32 KB, 500x531)
32 KB
32 KB JPG
Happy anniversary guys! We made it 3 years!!

https://desuarchive.org/g/thread/92419465/#92428057
>>
>>108479843
>>108480049
>gcc
Why would anyone use this piece of shit?
>>
>>108480150
And in the mean time we finished, like, one game?
>>
>>108480150
still can't tell if we're worse or better than /agdg/
hell, at least we're not /dpt/
>>
>>108473772
dove right into vulkan and honestly find it logical and even illuminating as how to graphics work on a deeper level
>>
>>108480615
What do you mean by a thread being worse? How are they worse?
>>
>>108480615
You're kidding, right? /agdg/ doesn't even get progress posts or on-topic discussion anymore, it's just a handful of spiteful anons baiting each other 24/7.
>>
After swapping out the handwritten matrix math with the native DirectX math library and adding a couple basic optimizations my stress test went from 280 fps to near 800.
>>
>>108480847
>stress test
What about a real world scenario?
>>
File: ygg-1774814004.webm (2.53 MB, 1920x1200)
2.53 MB
2.53 MB WEBM
Made this game cursor earlier with the best temporary sprites I could muster up for today
>>
>>108480615
/agdg/ is currently plagued by schizos and a spammer from basedjack party. One of which is obsessed with trannies and spams gore and wojacks.

>>108482266
I like it.
>>
>>108482325
>One of which is obsessed with trannies and spams gore and wojacks.
Acrolyte dev?
>>
File: AIE2.jpg (71 KB, 1129x843)
71 KB
71 KB JPG
Considering the fact that the AMD AI max has an npu that has the same flops as it's APU (which is roughly the same as a 4060), could it be possible to use it for something that isn't AI?
I looked at the MLIR and the architecture of the Xilinix AIE that's used in the NPU's and it seems like absolute hell and you need to think about the instructions and memory in terms of tiles and the various ways of communicating between the tiles.
I was thinking maybe since the tiles are laid out in a 2d grid, could that be used for a 2D map, like lets say a falling sand game? I wouldn't want users to NEED an AMD AI npu, but like what if I had a tile grid that was super coarse BIG pixels, but if you had an AI npu, I could make the pixels be 10x smaller or something.
I also saw llvm offloading, but I think it wouldn't work well with NPU's (at least AMD's, I have not looked at intel's npu), and NPU's are kind of neat because they "should not" affect the performance of GPU rendering as much as a compute shader / Cuda/ etc.
>>
File: 1462655383045.png (46 KB, 184x184)
46 KB
46 KB PNG
How do you guys get into the groove to work on your game, knowing it's going well, but being so drained from programming as a day job? It's the weekend, and I've slacked off so hard on the internet.
>>
>>108482343
Some guy who makes mods for games.
>>
>>108482592
Do a day job that isn't coding
>>
Anyone with a simple ui-layout kind of thing to spare?
I'm getting filtered.
>>
>>108483075
but I make too much money
>>
File: spaceinvaders.webm (1.5 MB, 804x623)
1.5 MB
1.5 MB WEBM
progress.

:D
>>
>>108483096
>>108482592
stop trying as hard at work or work on your game during work hours if you can get away with it.

i keep a pocket notebook with me while im at work and write down ideas for how to solve problems i'm having with my game, a to-do list for after work, and general ideas about what i want to do when i have more free time.
it's nice to have a physical representation of my game while i'm away from home. and since its a notebook with no backspace or anything like that, and all i can do is add to it by writing things down, it feels like a constant forward momentum even if its all in my head.

basically gotta work on it a little bit here and there during the week before or after work. then on your days off, the idea of jumping back in isn't as daunting, you'll already have things that are in-progress from earlier in the week, and things written down on your to-do list.

best way to get into a groove and maintain it is to keep things moving, even if you gotta slow down for 5 days of the week its better than some start stop bullshit.
then take a complete break once or twice a week where at most you only write things down. i use break days like that for writing down a more generalized to-do of things that need work, and then over the next week or two thats I'm keeping in mind when writing down stuff in the pocket notebook, game devving when I get home from work, and on days off.
>>
>>108483534
>then take a complete break once or twice a week
oops meant once a week, or every other week.
>>
File: 2026-03-29_22:32:06.png (301 KB, 640x480)
301 KB
301 KB PNG
How do you get video files smaller than 4MB?
I tried using OBS. it always looks like a blurry mess, and still comes out too big, even with 640x480 resolution and a length of only a few seconds.

I want to show my rolling field of Konas buzzing around in circles. They all follow the terrain, and turn their mesh to face the camera at all times. Also there are over a thousand of them, and the lowest the framerate has dropped is around 100.
>>
>>108479746
use ispc
>>
>>108482470
problem with using any npu is documentation and lack of hardware support in open source drivers. Its a scam since they advertise features you cannot use even with their in house tool chain. I am quite sure the npu will have the same problem as offloading to the apu the transfer cost itself makes it not worth it for anything complex.
>>
>>108483707
Record in high quality in OBS, run it through ffmpeg (or some frontend to it) to cut it down massively and fit under 4MB.
Getting realtime high quality yet small file size recordings usually isn't going to happen.
>>
>>108484419
yeah, this. real time encodes don't have the time budget for much optimization, but ffmpeg can squeeze it down and take as much time as it needs
>>
File: test.webm (2.12 MB, 1280x720)
2.12 MB
2.12 MB WEBM
Things look a little glitchy but I'm pretty sure its mostly because of the interpolated frames between update ticks. I've faced this kind of problem before where an interpolated frame setup doesn't take to kindly to teleportation.

Here is a basic animation test. You can adjust speed and do looping. Made with 5 translation key frames and 2 rotation key frames.
>>
File: 1766451398044430.png (449 KB, 1188x936)
449 KB
449 KB PNG
Not sure if it was discussed already, but the Slug algorithm for text rendering is now public domain. Kind of a big deal.
>>
File: kot.mp4 (1.81 MB, 854x480)
1.81 MB
1.81 MB MP4
>>108483707
i have a vibe coded .bat file that when i drag and drop a video file onto it uses ffprobe/ffmpeg to calculate a target bitrate based on the allowed filesize and do the encode
i'd post a video it actually made but it's all porn for another board, so here is a cat
>>
>>108484581
OH SHIT
>>
Why do people that write their own engines implement their windowing/input abstractions instead of just using SDL? Isn't that the boring part? Time spent doing that could have been used on rendering, math, physics, audio, tools, etc.
>>
>>108483208
AAAAAAAAAAH!
>>
>>108482153
My current game is nowhere near demanding enough.
>>
>>108484581
>not selling to the Rothschild patent troll
Based
>>
File: 1688388002119650.jpg (164 KB, 512x347)
164 KB
164 KB JPG
>>108474708
Holy fuck I'm old.
>>
>>108484581
sick if true. can't find that text on the site though
>>
>>108483534
>stop trying as hard at work or work on your game during work hours if you can get away with it.

that doesnt work like that
>>
>>108483534
That’s insanity. Don’t prioritize your toy project over an actual source of income. You’ll regret this.
>>
>>108484769
no one does this people just use GLFW or whatever else
>>
>>108477372
>Putting cube in world
I see the project is basically completed.
>>
>>108484769
>Why do people
they don't
>>
>>108484769
It’s such a simple part. No reason to use a library for that.
>>
>>108486417
knowing him, he will do the same thing again
>>
File: 1762010750926820.png (257 KB, 648x707)
257 KB
257 KB PNG
>>108486604
Autism and repetition and all that.
>>
File: 1583426351859.jpg (26 KB, 365x397)
26 KB
26 KB JPG
Oh yeah, I want you guys to recommend me books on physics engines. If you have any suggestions I'm all ears.
>>
>>108486614
do you ever plan on doing anymore more?
>>
>>108486775
I'm not sure I understand the question.
>>
>>108486811
will you ever do more than a spinning cube?
>>
>>108477372
>>108486614
>search the archives for name
>posts from early 2024 showing a quad being rendered
>2 years to upgrade to a 3D cube
Anon, you need to stop adding indirection and other bullshit and make game. This should take 2 weeks, not 2 years.
>putting a cube in world using CHICKEN
I have no idea what that stands for, but adding a cube should just be 1 function call with a a world pointer, a transform, and some identifier for the mesh.
>but what if I need to call it from another computer over the network via a scripting language in a VM?
Stop
>>
>>108487029
The first time I tried a Vulkan engine was when /gedg/ started. I went on a pretty big hiatus because of grad school and would post my other projects occasionally. They never really led to anything though.
>>
>>108487029
>2 years to upgrade to 3d cube
Eh last thing I remember on that engine was that I was able to generate a cube world and I was using lua scripts to alter the 3d generated world. I also had A* pathfinding on the cube world.
>>
What is a command buffer?
>>
Do u play your own game everyday and get hooked just like when you play other people's game?
>>
>>108487311
Yes. I’m bothered it distracts me from developing the engine further
>>
>>108487277
A queue you put commands into and the GPU consumes them as fast as it can.
>>
File: 1751539268608809.png (80 KB, 220x220)
80 KB
80 KB PNG
>>108487390
GPUs do vore now?
>>
/gedg/ is my new home from now on, not posting onto /v/ game dev threads anymore after today.
>>108487277
in Unity its a command you send to the gpu for rendering at certain parts of the pipeline.
so say you want to make a stealth game, and use a mesh for your shadows instead of something on the fragment shader, since its a realtime thing that mesh might need to be somewhat dynamic
so you could do shadow calculations for your atten, and then use the command buffer to push the instruction for how and when to draw the mesh for shadows.
>>
Post your /gedg/ icks.
>he made a scripting language for his 1 man project
>>
>>108487754
>>
>>108487754
>he's using some made up language like jai or zig
>>
>>108487754
>he thinks hot reloading is amazing
>>
>>108488234
reloading is a dish best served cold.
>>
>>108487754
>his game only works on windows
>>
>>108487754
>using godot unironically
>>
>>108488234
What’s wrong with hot reload?
>>
>>108487754
>ECS
>>
>>108487754
What the fuck is wrong with a custom tailored debloated scripting language?
>>108488234
>hotreload denier
Opinion discarded
>>
>>108488269
any game that works on windows works on linux nowadays unless you're EA and you purposefully hire 200 developers to tweak your anti cheat to block linux users
>>
>>108484581
>waste time and money to get a patent
>give it away
???
anyway, algorithm patents don't mean shit around here
t. EUfag
>>
>>108488636
Linux autists just throw a hissy fit over the idea of running games through a translation layer that at most takes 1% performance away
>>
>>108488658
im pretty sure the translation layer makes games more performant on linux than their windows native version nowadays
>>
>>108487754
>OOP
>>
>>108488875
this
>>
>>108488979
We don’t want ur kind here
>>
>>108488991
now calm down, anon, he ain't hurtin' nobody
>>
>>108488991
whatd i miss?
>>
>>108488402
>>108488628
abi boundaries are too strict, you cant add a field to a type without it all breaking. the only thing a hot reload can work with is purely functional
>>
>>108489786
You can, by only ever adding to the end and having something else to indicate what's there (e.g. version number, sizeof struct), but that shit is way way too much effort for hot reloading just within your own program. That's the kind of thing kernel interfaces do.
>>
File: 1750324892050179.jpg (122 KB, 482x427)
122 KB
122 KB JPG
>made a full 3d engine in flash back in the day with 0 knowledge or experience
>can't even find the motivation to make a 2d engine in raylib now
>>
Is it more "correct" to have one shader that can do many things, or multiple shaders that can do one thing each?
so, would it be fine to have one shader for meshes and terrain, one for particles, and one for UI? Or is it better to not have to keep switching between them every frame?
(OpenGL by the way)
>>
Do you all hate Bevy & Rust?
>>
>>108490637
I think large engines get on the level of doing a pipeline per material, which is why modern games can spend literally dozens of minutes compiling shaders on first load. It's not like they actually wrote that many shaders by hand though.
But then you have the complete opposite end of the spectrum, where you have ubershaders, but the original example of that was to fix shader recompilation stuttering in an emulator.

Really to say, I wouldn't say there is a "correct" solution. Just having a handful of shaders like you're describing is perfectly resonable and sound. I wouldn't say that generating a gorillion shaders is "more correct".
>>
>>108490637
>correct
ngmi if you're even thinking like that

and yes, you should use an ubershader unless you have a good reason not to, branching in shaders doesn't matter if every pixel currently be processed takes the same branch anyways
>>
>>108490710
no wonder modern games are so bloated and buggy.
>>108490723
I put it in quotes for a reason.
I'm just trying to figure out some generally accepted wisdom from people who know how the API works.
What would constitute a "good reason" in this instance then?
I'm not going for some sort of graphical marvel, so I don't think I have to worry, but I'm still curious.

I'll probably just stick to my ubershader in this instance.
>>
>>108485687
It depends. If it's an average software dev job, then it should be fine and can actually boost your productivity with work as well. Because your brain gets more stimulated this way and helps to not be sleepy all day.
>>
is there any free hosting that can handle something basic online gameplay for a small number of users?
>>
>>108491094
you can get a low-end VPS for like $10/year. check offers on lowendbox.
>>
>>108491094
>free
If there were then indians and chinese would make millions of accounts and set up crypto miners until the entire service collapsed.
>>
>>108491152
This, rip heroku my beloved
>>
>>108491152
what about some kind of proxy service where I can host everything myself but not have to directly give out my personal ip?
>>
Claude in computer control mode is ridiculous. I have never made games so fast
>>
>>108491463
you can use a VPN with port forwarding
>>
>>108491463
Nothing free that'd be even remotely reliable and isn't somehow malicious. The only reason you can even get a free email these days is because the provider farms your data and sells it.
>>
>>108490637
one shader that can do many things aka uber shader. That’s why iD software does for Doom and they are able to keep pipelines down to less than a couple hundred. Unreal has pipeline explosion because making a new material means making a new hlsl shader which means making a new pipeline. So there is technically an infinite number of potential pipelines. Ideally you want to be not totally dynamic like Unreal and keep as much logic as possible packed together hence uber shaders.
>>
Have any of you guys implemented multiplayer in one of your games? If so, what are your standard practices? Do you implement listen server or dedicated server? Saving to disk or to database? And do you use a modern replication module?
>>
>>108491982
Is also like to hear more about this
>>
File: john_doom2.jpg (127 KB, 732x462)
127 KB
127 KB JPG
I am happiest when working on graphics code and C++. When I get into a flow working on some stupid little graphics or linear algebra problem, I feel invincible. But the entire goddamned world keeps getting in the way of that. Work problems, life problems, family problems. Everything is trying to keep me from doing the one thing that makes me happy
>>
>>108492656
I sorta know that feel, but I'm a bit of a math retard... for now.
>>
File: Interiors - pg1.jpg (347 KB, 1236x1604)
347 KB
347 KB JPG
>>108492687
So was I. 1.7 GPA in high school, failed my way up through community college, transfer to and dropped out of big boy college after getting filtered by calculus 3 times.

But somehow, when I need math for graphics/physics/engine dev, it's easy. I guess calculus was never hard, it was just hard to convince myself to go to class
>>
File: 1733332492822.png (210 KB, 400x400)
210 KB
210 KB PNG
>>108492656
Why you post Romero and not Carmack though. Romero is le lazy art/mapping dude, Carmack is turboautism engine/graphics programmer. Carmack fired Romero because he was too lazy and had too much fun at workplace. Romero is cool af too, just less relevant to your post.
>>
File: RS_Romero.jpg (97 KB, 655x871)
97 KB
97 KB JPG
>>108492731
The cult of John Carmack vastly overstates his technical achievements. Carmack was narrowly focused on visible surface determination during quake's development - they hired other god-tier engineers who were lost to history to handle other graphics and networking tasks. BSPs were not something Carmack invented, just appropriated into the world of real-time rendering.

In Romero's telling, Carmack basically started to lose his mind during the development of Quake because he was cracking under the pressure. Carmack prides himself on his "work ethic" but the reality is that he was struggling heavily with VST because he wasn't sleeping, wasn't exercising, and survived on a diet of pizza and diet coke. As a result, he began lashing out at his coworkers, who had to put up with his shit because he was their engine guy. You can get a lot more thinking and working done when you drink water and get enough sleep.

Also Carmack is an AI grifter now
>>
File: dhs025.jpg (110 KB, 809x547)
110 KB
110 KB JPG
>>108492731
>Carmack fired Romero because he was too lazy and had too much fun at workplace

This is a falsehood from "Masters of Doom". Romero was forced out because Carmack thought that Romero should have been spending 100% of his time on quake instead of project managing Heretic, Hexen, Master Levels, Final Doom, and Dwango. Carmack believes that companies should basically only make one thing, while Romero wanted to grow the company and diversify their income. I don't even think Carmack disputes this. Masters of Doom contains about a dozen factual errors and this is one of them.
>>
>>108491982
I did implement the technology for multiplayer, tested it with multiple simultaneous clients, but my current game is not meant for multiple players.
>>
>>108492775
>Also Carmack is an AI grifter now
What's he actually working on? He was talking about it like it was the way of the future on that dudes podcast, but I i never actually looked into it.
>>
>>108493150
Last I heard he was training AI to play atari games or some shit. How the mighty have fallen.
>>
>>108494068
If I had his money, I'd probably do stupid shit too. People will throw tens of millions at whatever he wants to waste his time doing, so he might as well get real weird with it.
>>
>>108472901
I found a WIP indie game someone shared with me 12 years ago. It was a fully playable level 1 demo and they gave me the project files, not just an executable. I searched for it online and it looks like they never made a full release.
Would it be improper to try building onto it as a learning project?
>>
>>108494374
post the files
>>
>>108494374
I wouldn't touch it. there is a fuckton of freely available open-source projects to build on. it's pointless to waste time on some obscure project with uncertain licensing.
>>
Unity and UE5 are dead
S&box is the new king of engines
>>
File: 1752235603954429.png (37 KB, 715x150)
37 KB
37 KB PNG
>>108494543
based gaben
>>
>>108490496
>made 3d engine in flash
>didnt simply use shockwave/director
>>
>>108494543
>>108494636
>source 2 engine
I'd imagine its 2D is pretty weak?
>>
>>108494543
Nobody would use Source 2 even if they released the full thing. It would be at least a decade behind Unity and Unreal in terms of tooling, which is what really matters to developers.
>>
>>108494839
I'm not sure what you mean, Source 2 is just the backend engine, S&box is a whole IDE just like UE5/Unity, people have already been making games with it for years but you weren't able to release games standalone
>>
>>108494873
That's even more of a non-starter because it's going up against Roblox. Fortnite tried it with a concurrent player count that was a third of Steam's total online users at the time and it did nothing to even shake Roblox.

Aside from that I would steer well clear of anything that Garry Newman has his pudgy fingers in. He's an insecure faggot who will fuck over his own product to spite people he's butthurt at.
>>
>>108494917
are you not able to read or something?
>>
>>108494917
>Roblox
???
I don't think you're following m8, they negotiated a new license with Valve and now you can build standalone games and sell them completely royalty free
>>
>>108494936
>>108494950
There's nothing new here, it's just another shitty toolset like Roblox and Fornite Metaverse (or whatever the fuck it was called), not even a proper engine (I highly doubt you get access to Source 2's source code). All you'll do is waste time on yet another novelty that will bring no fruit.

Stop trying to find the latest "tech stack" and get back to actually working on your engine, there's no magic solution coming to save you from having to work your ass off.
>>
>>108494962
concession accepted
>>
>>108494962
Bruh, take your meds, it's a full game engine just like UE5/Unity except royalty free
I'm sorry that this is happening to you
>>
File: 1774649779026096.png (218 KB, 630x164)
218 KB
218 KB PNG
>>
>>108494970
>>108494983
he's a mad pedot user who just realized that he wasted years of his life on a deadend engine
>>
>>108494983
Great, and you had Unity and Unreal available for over a decade before this. How much did that help you?

Aside from this being a thread about engine development, there's also no point in chasing new tech like you do. Obviously you think this will help you create your game or whatever, but there are established solutions that have been available for years and you could have done whatever you expect to do in S&Box there and be years into development by now. You didn't do it, they didn't help you, you do not need yet another layer on your "stack". You need to stop looking at this kind of magic solution bullshit and start working on your actual fucking game.
>>108494995
I like how you assume I'm working in fucking Godot in a thread specifically made because people working on their engines didn't want to go to /agdg/.

And this is exactly the kind of mindset that is pervasive in /agdg/: "just one more engine/plugin/LLM/workflow/DCC/whatever and I'll finally make my game!". Shut up and start programming, retard.

Also
>royalty free
Are you allowed to release it outside of Steam? Because otherwise it has the by far highest royalties of any engine/framework/toolset.
>>
>>108495044
I'm getting second hand embarrassment for you
Just take the L bruh, it's an anonymous image board, nobody will know
>>
>>108495044
>/agdg/
And yet here you ae acting like a little bitch because you have some kind of unresolved trauma related to Garry's Mod.
>>
i took a break from the 3d engine/game im working on to work on a 2d one and man i am making so much more progress so quickly , i see why people do this
>>
>>108495068
>>108495099
Good luck with your endeavour in S&Box, it didn't work out the other 15 times but you've got a great feeling that this is the one!
>>
>>108494543
can I do standalone builds right now? want to test the performance
>>
>>108495044
>made because people working on their engines didn't want to go to /agdg/.
Nobody except schizos want to go to /agdg/.
>>
>>108495222
the 28th
>>
>>108495044
>Great, and you had Unity and Unreal available for over a decade before this. How much did that help you?
what a weird thing to say, this sounds like a projection of your own failure to make a game
>>
File: 1750631803883423.jpg (177 KB, 1024x1024)
177 KB
177 KB JPG
>noooooo source2 isn't a real engine it's just roblox!!!
>>
>>108495739
kek this dude's been seething for the last 2 hours so he had to come back and make a passive aggressive post without replying
>>
>>108495739
nodevs do be like that
>>
>>108494068
It's not that bad idea to use AI for NPC AI imo. With some cool special-purpose models, NPC AI might be both way more flexible that average but at the same time very efficient compared to LLMs.
>>
>dude just use local llm in your games
>"get lamp"
>*gpu screams and bursts into flames*
>>
>>108496476
>Minimum system requirements: RTX 5090 x 2
>>
>>108496476
Do you think Carmack uses LLM? Maybe he's building custom model for NPC AI.
>>
>>108496476
Soon you'll be able to buy lab-grown human neurons on a usb stick for AI acceleration.
>>
source 2 kind of sucks ass to use as a developer. s&box makes it easier but it still fucking sucks ass to use. just a really terrible toolset
>>
File: OilyVulture.webm (3.98 MB, 1406x960)
3.98 MB
3.98 MB WEBM
Got particles spawning on meshes now. Honestly thought it was going to be more difficult. Just need to crank out a few anims on this here vulture and we're in business.
>>
>10 million wishlists
I made it.
>>
>>108496910
crazy with all the games valve makes you think they'd have great tools for it
>>
>>108496921
nice, reminds me of that one sm64 level in the sewers
>>
I've heavily researched that market and I've found that all people want is mine slop
>>
File: 1771826085428362.png (11 KB, 224x288)
11 KB
11 KB PNG
>>108497108
how could one game be so far ahead of its time?
>>
>>108497108
>>108497117
i HATE diggers
>>
>>108497108
The become a shover seller thing wasn't a joke
>>
File: ideaguy.jpg (36 KB, 372x372)
36 KB
36 KB JPG
incremental mineslopper tower defense liminal horror anomaly crafting with tactical rpg elements
>>
>>108497192
You fool, you forgot the card game. You can't ADHDcore without a deckbuilder.
>>
File: 1773054409030391.png (892 KB, 1280x720)
892 KB
892 KB PNG
Is there anything worse than when you're playing a game then all of a sudden this shitty minigame is forced upon you?
>>
File: packneel.gif (4 KB, 227x188)
4 KB
4 KB GIF
I'm slowly cobbling together a game in C with the standard libraries, because I enjoy reinventing the wheel
I'm using a sparse set ECS for all of my data, but I'm not sure how best to "disable" entities. When the player goes to the next level, should I systematically delete the data for the enemies from the previous level? Flag the data as disabled and have the systems skip over it? Copy the player's data into new memory and free the rest?
I don't have a CS background and am learning as I go, this is my genuine stupid question and not bait
>>
>>108497355
I mean yeah you could just have a list of active entities or something like that. If its relevant to the component data you can also store it there (like maybe you want to do a physics update on a physics component but don't want to draw a mesh component)

Depending on the situation you can also just delete the relevant component. Some of my components are just indexs/keys into other data structures.
>>
>>108497355
>C
>ECS
2 strikes, you're on thin ice
>>
>>108497377
>If its relevant to the component data you can also store it there
I think I've read about implementations that use the first bit of the id variable as a flag for whether the component is enabled or disabled. I'll see if I can dig that back up
>>108497381
I did it initially because a stranger on the internet told me ECS good, now that it's starting to take shape I actually quite like it
>>
>>108497355
Tear down everything, start from scratch on level load.
>>
>>108497320
The main problem with forced minigames to perform repeated tasks is that the minigames rapidly become the core gameplay loop. If you have to solve a pipe maze every time you open the door, congrats, you've basically created a pipe maze game that's frequently interrupted by unrelated gameplay.
>>
File: 1746436286698237.png (97 KB, 970x561)
97 KB
97 KB PNG
don't forget to thank your system notifications
>>
>>108497595
It depends. Lockpicking in TES doesn't feel like this, for example.
>>
What engine do people use to make the psx low res garbage?
>>
>>108498564
Godot, Unity, even UE5.
Pseudoregalia is UE5 for example...
>>
how's crossplatform dev on linux?
>>
>>108498744
From the sound of it Linux users can't go 2 days without a getting pwned by malicious packages in their package managers
>>
>>108494543
>>108494636
thisis a game changer
>>
I need a ton of help with this nonsense.
>Try to make dolly system with multiple points
>Works-ish
>Yet whenever you get into the last "Segment" you just cut back to the second camera.
>>
>>108499307
I did some more testing and now I'm even more confused
>>
>>108497355
Create a structure to function as an ECS world. Keep all the storage for entities etc in that structure. This also means create_entity, get_entity, whatever will take a world pointer as well.

Now just delete the world and create a new one when you transition to a new level. Copy over the entities that you want to keep (e.g. player).
>>
File: HUH (1).mp4 (1.53 MB, 1280x718)
1.53 MB
1.53 MB MP4
>>108499474
>>108499307
https://pastebin.com/7w3jVZHj
Like look at this shit. I genuinely cannot make heads or tails of it
>>
>>108497381
ecs is good
>>
>>108499505
try asking claude a few more times
>>
>>108497355
>sparse set ECS
whats the point of this?

>>108499556
>ecs is good
why?
>>
How bad is it to have each model with it's own VAO, VBO, and EBO in OpenGL?
How many of these would verge on the "too fucking many" zone?
>>
>>108499563
google ecs
>>
>>108499582
I know what it is
>>
>>108499577
each model normally has a VAO and VBO or multiple VBOs
no idea what an EBO is
>>
>>108499586
question answered
>>
>>108499598
I didn't ask what it is retard
Sparse set just takes away any benefit of ECS, localized data
>>
>>108499589
Element Buffer Object.
And you can set one VBO to store as many models worth of data as your video ram can hold. I don't know what you mean by "normally has" in this case. it's all decided by the programmer.
>>
>>108499604
im not the sparse set guy though
>>
it surprised me to learn that retards have apparently retconned what "mob" is short for in their headcannon
thanks for reading my blog
>>
>>108499606
The most common way to do this is for each model to have its own VAO, VBO and EBO
>>
I'm developing my game on an RPi 2 to ensure it can run on a potato.
>>
I'm developing my game with mesh shaders and descriptor heaps to prevent poor people and third worlders from playing it
>>
>chatgpt, translate my game to traditional chinese
you're not just throwing away the largest market on the planet, are you?
>>
>>108499655
bruh, you think I have time to create a bitmap font that includes every kanji?
>>
>>108499614
Yeah, I thought so.
I'm trying to work out a bug where the game hard crashes after a minute or so depending on how many models I load in. It happens in about 20 seconds if I load up around 5000 character sprites. (it's a 3d game, they're just billboard sprites)
ChatGPT is insistent that I should try loading everything into a single buffer to use in snippets, but then kind of didn't have a good answer for what I should do when I want to remove some data from the middle of that buffer.

What point would it be wise to start consolidating? (I'm working on a way to have all billboard sprites use the same model object, but I'm still wondering what's going wrong.)
>>
>>108499628
You know you can use like profiling tools and stuff so you can dev on a real PC and still make sure your performance is within whatever spec you want, right?
>>
>>108499705
I thought you were talking about a 3D game
If you're drawing sprites you might want to take a different approach
>>
>>108499716
I already said it's 3d and I'm working on a new approach.

The new plan I'm thinking of now is to have every cell have one set of buffers for the base terrain, one for the buildings and other static fixtures, and one for the dynamic entities like interact-able items and characters. One cell will be around 512x512 meters since being open world, I can't load everything in all at once.
>>
>>108499740
Oh I misread
I don't think it's normal to crash after 20 seconds if you have 5000 models loaded, sounds like some other bug
Sharing buffers for models isn't really necessary either
>>
I make claude talk like an anime girl and call me onii-chan
>>
>an unbounded descriptor table still needs a bounded descriptor heap
that doesnt sound very "unbounded" does it?
>>
>>108499985
>Vulkan



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.