Everyone is saying AI will take your job, but think about it another way - programming has become so complicated that draw a simple triangle on the screen is now 1000 lines of code and you would need hours of digging through obscure specifications just to figure out what any one of these lines does. That's why video games take 10 years to develop now. We need something to speed things up, you can't just spend a year writing boilerplate for vulkan before you render a single polygon to the screen. But wait... as soon hardware becomes more complicated than the human mind can handle and coding becomes more of a chore than development time allows, we magically have AI. Coincidence? I think not. There's really only one possibility, we live in a simulation and to continue being productive the matrix has gifted us this tool without which everything would stagnate. It all makes sense really.

>>105840008 (OP)
>That's why video games take 10 years to develop now.
Absolutely not. 99,9% of games are more or less assets dragged and dropped in Unreal Engine now.

>>105840008 (OP)
AI sucks at coding.
It can only do the most basic shit and even then the code needs to be reviewed by a real person.
It can't to complicated programming.
People who program using Vulkan are highly specialized and AI won't help them.

>>105840008 (OP)
nobody rawdogs Vulkan, come on

>>105840008 (OP)
low level stuff has been more complex to program than high level stuff since the dawn of time so no you're just retarded and schizofrenic

>>105840008 (OP)
You can still use OpenGL, but Vulkan made a lot of choices that make it possible (compared to OpenGL where it's impossible) to write extremely performant code that uses 100% of the GPU.

>>105840008 (OP)
AI created games coming up

>>105840008 (OP)
It always took 1000 loc to draw a simple triangle on the screen.
Just that the code wasn't accessible to you before vulkan and were supplied by the graphics driver.
If you don't want deal with it don't. Either still use opengl or something like magnum engine that still gives you full control but takes care of the boilerplate for you.

>>105840090
I rawdog Vulkan. 1.0 even.>>105840008 (OP)
>>105840008 (OP)
1000 lines is nothing. AI is a glorified content aggregator, it does poorly at doing specific things. coders are safe, creatives are going to learn to make models based on their style and use it as force multiplier.

I swear, AI fags are like crypto cucks.

>>105840008 (OP)
>programming has become so complicated that draw a simple triangle on the screen is now 1000 lines of code and you would need hours of digging through obscure specifications just to figure out what any one of these lines does
Is this true or is it a no-coder take?

>>105844781
It's true but it's a midwhit take.1000 loc is nothing and once you get a polygon rendered you're already half way to rendering entire models and environments. Needing to write more boilerplate code isn't more complicated, it just gives us more control.

>another AI schizo thread

>>105844871
So is it 1000 LOC to draw a triangle, or 1000 LoC to set up the grqphics environment to start drawing anything you want to the screen?

https://vkguide.dev/docs/chapter_2
It's only a few hundred LoC.

>>105844952
The latter pretty much.

>>105844952
the latter.
And thanks to extensions, you can entirely skip a lot of the shitty 1.0 API for even less boilerplate starter code.
https://github.com/SaschaWillems/Vulkan/blob/master/examples/dynamicrendering/dynamicrendering.cpp
VK_EXT_descriptor_buffer is a nice one
https://github.com/KhronosGroup/Vulkan-Samples/tree/main/samples/extensions/descriptor_buffer_basic
Vulkan is starting to trend towards less explicit abstractions, which is largely what 1.0 was designed around, and why 1.0 triangles were so tedious to render.

>>105840008 (OP)
Solves your problem

My OpenGL game doesn't have this problem

>>105845066
It can't even solve its own problem and be released.

>>105845066
Ah yes, the superior W*b software.

>>105845094
It is released it's just firefox is lagging with adding it to their browser.
you can use the C++ Dawn or the wgpu Rust implementations right now.

>>105845116
works on my machine

>>105845187
that's great, but the entire point of web software is that things "just werk". I'm not even using an ancient freetard firefox fork either.

>>105845201
>but the entire point of web software is that things "just werk".
Javascript ended that years ago.

>>105845021
>2025
>descriptors
>vkGetDescriptorSetLayoutBindingOffsetEXT
that's like opengl 2 when you had to manually query uniform offset. as always, vulkan people can produce only garbage.

>>105845148
>wgpu
>issues 906
it will never be released. it's unstable, unusable, slow garbage.

>>105845540
then you learn about BDAs, and can then work in gpu addresses.
https://github.com/KhronosGroup/Vulkan-Samples/blob/main/samples/extensions/buffer_device_address/README.adoc

>>105845556
doesn't work with textures. why can't they make it like opengl where you just need to call 2 functions, instead of writing 1000+ LOC to manage descriptors yourself?

glGetTextureHandleARB()
glMakeTextureHandleResidentARB()
it's that simple in opengl.

>>105844781
It's false. On Linux, /dev/fb0 gives you access to the main frame buffer.

>>105845592
>doesn't work with textures.
it does though, what do you mean?

>>105845555
custom shader language and compiler is what killed webgpu. it's just too much to implement.

>>105845630
why did they drop spirv anyway?

>>105845592
Bindless textures still exist afaik, not that i ever had to use them.

>>105845630
>>105845634
you can use spirv (at least on non-web) since it's in the implementations.
but it isn't in the spec.
also I'm not sure if there's compute shader support but I have used fragment and vertex shaders doing hlsl -> spriv with dxc on webgpu.

>On June 8, 2016, Google showed "Explicit web graphics API"
lol, webgpu is as old as vulkan, and still isn't finished (even in chrome it works only on desktop)

what a failure

>>105845843
If anyone wants to know why there's a good explanation from the SDL people.
https://moonside.games/posts/layers-all-the-way-down/

>>105845634
WebGPU must be safe, to not crash the browser or do other nefarious things. spirv isn't safe.
They compile wgsl to safe spirv with inserted checks and so on. This is a mess.

>>105845871
I wonder how vulkanSC users handle shaders

>>105845116
Just wait two more weeks. I mean literally. There's a new Firefox stable release in two weeks that enables WebGPU support.

>>105845909
I use Floorp.
Apparently v12 released recently, but no idea if it enables it

>>105845066
this shit's already technologically dated by a decade

I think you're on to something but you jumped to a conclusion there. I noticed a similar phenomena with Rust - the lifetime management is a bit much for humans, and the compiler errors are very verbose, basically the perfect language for AI. Can I prove it was intentional? No. Can I explain the timing? No. There are no coincidences, but that doesn't mean we can just "connect the dots" without errors. That's the problem with conspiracy theories. Truther-ism is better. Just tell the truth and seek truth, eliminate assumptions and jumping to conclusions.

>>105840008 (OP)
Its true that real graphics programming is done by PhDs in that field, usually those work for game engines like unreal. Game dev is not computer graphics. No game dev implements anti aliasing, mipmapping, supersampling, etc. themselves. Game devs write game logic and they let the engine do the rendering. However most (game) devs are retarded indians who can't even tick a box to turn on some of these amazing features.