Vulkan Forest Scene

[topaz cove]

oh OpenGL

[solid jackal]

im definitely doing opengl next

[thick edge]

Hehe

[topaz cove]

its midnight so I was confused if demongod was sayin graphics library

or somethin

[uncut tusk]

lmao

[topaz cove]

they waterboarded my cohort with mandatory vulkan in sophmore year in undergrad

urm... dunno why i got to take it before I completed OS class 😅

[solid jackal]

I know too much about the API but not enough about techniques

[uncut tusk]

Literally me

[thick edge]

Idk shit about techniques I just invent everything as I go

[uncut tusk]

I just read papers

a lot of them

[topaz cove]

I feel like this is a good rule of thumb in general because u can bruteforce and tryhard above ur skill level and still kinda develop but then u get strange gaps in ur foundation u gotta go back n fix anyways and its more stress for less gain

[uncut tusk]

too many probably

[topaz cove]

It's like urm, lifting and stuff

You could probably egolift with bad form if u really force urself

but u risk just hurting urself even if that gives u gains in the short term

[solid jackal]

yeah

I lift too btw

lmao

[uncut tusk]

Actually how the hell do you know if you're in that zone

[topaz cove]

wu wei or some shit

[topaz cove]

I think its easier to tell when u aren't but idk any coding analogy for it off top of my head

but it'd be like trying to learn how to read a new language by bruteforcing a dictionary and subtitling a full movie in it straight away like u would be spending 150% of ur effort for like 5% progress

or like trying to grind intergration before differentiation or somethin

the easiest analogy is probably seen in fitness but i think the same stuff applies to the brain too, although this i dont have any citations for it

[thick edge]

If you only need gentle guidance from articles and can play the rest by ear

[topaz cove]

Where if u try and force urself to do certain exercises or lift certain amounts before ur body physically cant over n over u just arent getting as much gains ig

[thick edge]

If you're copying anything verbatim you're not there

[uncut tusk]

What if I'm copying maths

[topaz cove]

oh hell yea I am in that zone rn

[uncut tusk]

Because I'm sure as hell not going to reinvent the wheel in maths

[thick edge]

If your goal isn't to learn the math then copying the math is ok I guess, it depends on what your goals are

[topaz cove]

its actually kinda nice to know that u encourage it. I kinda always took if I have to 'try too hard' beyond that level of being able to understand from an article or description and do the rest myself that I don't have the pre-req foundation yet but i didn't know if that was wisdom or laziness

[uncut tusk]

My goal is to learn, understand and apply it

but the formulas themselves I don't really need to reinvent

[thick edge]

I guess it depends on what we're talking about

Are you talking about like light transport stuff

[uncut tusk]

Fourier, "higher" analysis (change of variables theorem, multidimensional integrals etc.) and some formulas that just "exist"

Like idk, I'm taking a "top down" approach where I see the stuff that I need for my ocean sim and then break that stuff down part for part until I reach the stuff I understand

[topaz cove]

i can relate to top down approach vibe

wait no i think im bottom up? but i relate to 'i only touch the stuff i understand'

[thick edge]

It all depends on your goals, if your goal is to be able to invent stuff like the Fourier ocean then you'd probably need to study those directly out of a book since wielding those tools creatively takes a lot deeper knowledge, but if you just want to be capable of implementing effects like the ocean then it's fine to just learn only what you need

I'm spoiled because I studied physics so pretty much everything in graphics is right in my wheelhouse and I can understand it from first principles just by skimming for concepts

Learning that shit on your own is hard man

[uncut tusk]

I kinda want to learn that shit to understand it enough so I'm able to invent stuff as well

but I learn best when I also have some practical application

[thick edge]

For sure

It just takes time

[uncut tusk]

Yeah definitely

[thick edge]

Are you in school?

[uncut tusk]

University

[thick edge]

What year

[uncut tusk]

~2 semesters (including this) until I write my thesis

[thick edge]

I take it you didn't have a multivar/vector calculus course then?

[uncut tusk]

I can technically take one but only get access to the lecture afaik

[thick edge]

Did you take linear algebra

[uncut tusk]

no exercises and definitely no grading

yes, was mandatory

[thick edge]

Did you take differential equations

[uncut tusk]

linear algebra, abstract algebra, some DEs, some calc

[thick edge]

Abstract algebra but no vector calc lol interesting

Is this for a CS major

[uncut tusk]

ah wait

some vector calc

but not indepth

we did jacobian etc.

but that's about it

[uncut tusk]

SWE, which at my uni means 90% CS

[thick edge]

Yea

[topaz cove]

holy shit my CS feels like 90% SWE

[thick edge]

Wait you did Jacobian but not even multivariable integrals?

[thick edge]

Mine was 90% computer engineering which is the best way

[uncut tusk]

we did partial derivatives

[thick edge]

Hmm ok

[topaz cove]

mayb intellectually but for early career I am seeing the benefits of that 90% SWE approach in how my clasmates and stuff are gettin internships left n right

alright time to sleep

have a good one u folks

[uncut tusk]

you too

[topaz cove]

also @solid jackal is in high school? proud of him fr fr

i learned to code like when i was old enough to drink 🙃

[uncut tusk]

but yes

[topaz cove]

(drinking age in Singapore is 18 tho)

[uncut tusk]

three semesters of theoretical cs, part of which is abstract algebra

it's also where most students that do drop out or fail reach that point

[solid jackal]

thank you 🙂

sunk cost fallacy best fallacy tbh

and nah im only 16 so i still couldn’t drink

[uncut tusk]

I can only laugh in German here

well we're allowed only the light stuff at 16 but still

[solid jackal]

lol

alchohol seems overrated tbh

[uncut tusk]

Nah, it is pretty good if you want to clean electronics

[solid jackal]

oh yeah

ISOPROPYL NUMBER 1 BABY

you aint a real one if you never had dis befo fr

[uncut tusk]

I literally cleaned off the thermal paste with vodka once lol

Didn't have anything else at hand at the time

[solid jackal]

oh lmao

i wonder if that other type of alchohol works

hydrogen somethingide

the stuff in the brown bottle

[uncut tusk]

there's a lot of brown bottles

you need to be a bit more specific

[solid jackal]

hydrogen peroxide

im not going to test it lmao

[thick edge]

That's not alcohol

[uncut tusk]

That's also not something I'd willingly put on my CPU either

[thick edge]

Haha

[thick edge]

Ok I had to look that up actually because I was pretty sure but then I had a crisis of faith when I realized that there is definitely an -OH group in hydrogen peroxide lol

But it has to be bound to a carbon atom to be considered an alcohol

[solid jackal]

bud thinks hes the knower

my pfp is atrocious

[thick edge]

16yo aren't you taking chemistry this year in school

[solid jackal]

this next year i think

i wouldn’t expect to learn much though

i happen to attend a school not well known for being good at anything

[thick edge]

Ah lol

[solid jackal]

I changed my grass texture, added some simple lighting, and got a better sky color

this looks like something from windows 95 lol

[thick edge]

Reminds me of mount and blade

I like it

Ship it

[solid jackal]

needs shadows lol

[thick edge]

Debatable

My games don't have shadows and they look good enough

I'll probably add them at some point but I haven't really cared thus far

[inland flax]

skipping shadows is a good way to have ez perf 🙂

[solid jackal]

i messed up

matrix multiplication confuses me so I just change it around until it works 🤣

[thick edge]

Lmao

[solid jackal]

well anyways i added some randomness

[verbal vector]

shadows are overrated. bad shadows are distracting, good shadows are hard to implement.
project some shadow disks under the trees and call it a day

[solid jackal]

alright fine

next is multisampling ig

[thick edge]

Isn't that just a flag you set

[verbal vector]

MSAA + transparent leaves = sadness (moreso the leaves aren't affected by it)

[thick edge]

Yea rip

Not an easy problem to solve

I'm trying to remember how I do it

I think I just don't lol

Or maybe I use purely geometric foliage

[verbal vector]

the leaves should already be filtered from the texture sampler to some extent... but this texture isn't doing any favors

I wonder if there's something wrong with the shader. my pixels are huge but come out smooth.

[thick edge]

Idk what I'm looking at

[verbal vector]

oh a zoomed in grate texture

[thick edge]

Ah lol

Is this with discard?

[verbal vector]

yeah

but I discard <= 0, not 1 which is what I think their shader does?

[thick edge]

How is it so smooth

Do you have antialiasing

[verbal vector]

I'm using a linear point sampler shader intended for pixel art

[solid jackal]

oh uh

im discarding at == 0

[thick edge]

Very pretty tbh

Ship it

[solid jackal]

nah

still feel like I haven't learned enough from this project

[thick edge]

Lol I'm joking when I say that

Mostly

[solid jackal]

lol

what do you think I should do next?

[thick edge]

The point I'm making is that that level of graphics is good enough for most development purposes

Idk it's sort of up to you, you need to find something motivating

What kind of games do you play

[solid jackal]

nah I mean what next for this project

UI maybe?

and I definitely want to do blending

[thick edge]

Yeah that's what I'm talking about too

What kind of games do you play

[solid jackal]

idk

I really like cave story but that's pretty primitive in terms of graphics

some FPS games

[thick edge]

Graphics should serve a purpose

Make a 3D cave story scene or something like that

Imo just making pretty pictures for the sake of making pretty pictures is slightly pointless, that's what raytracing is for

Vulkan and GL are for making interactive graphics

[solid jackal]

right

I mean this render engine I've made isn't really great to build anything on

[thick edge]

Why not

[solid jackal]

a ton of hardcoding

I guess I can fix that but still

[thick edge]

All my games have tons of hardcoding

Maybe not positions of things

But graphics are 100% hardcoded

Generalization in game graphics is generally useless

Unless you have a team of renderer devs to make an in house engine or something your graphics will be relatively hardcoded

Hard coding is how Vulkan is designed to be used, even if that hard coding is organized in config files or something

[solid jackal]

hm

[thick edge]

If you don't want to tailor your graphics very closely to your application there isn't much reason to use it over GL or something

[eternal dune]

Hardcoding is great for hobby projects but it's not how it was designed to be used

[thick edge]

It depends on what you're talking about

[eternal dune]

Keep hardcoding stuff until it works is good tbh

[thick edge]

Pipeline permutations are basically intended to be hardcoded or otherwise specified explicitly

Not doing that is why you have shitty modern engines that have to compile millions of pipelines for no reason

[eternal dune]

You can do specialization constants and uniform branching for that

[eternal dune]

Ye true

[thick edge]

https://zeux.io/2020/02/27/writing-an-efficient-vulkan-renderer/

Ctrl f for ahead of time compilation

[eternal dune]

Huh I somehow missed this blogpost from Arseny

Personally I think that the perfect solution is to have ridiculously large pipelines

Instead of many pipelines compiled ahead of time

[thick edge]

There are many different solutions for sure

The point of Vulkan is to let you tailor everything to the application and hardware

[eternal dune]

Yeah each solution has pros and cons

[topaz cove]

Unreal Engine games like Hogwarts Row be like

[inland flax]

hard coding isn't very real imo

code is meant to solve problems, just a question of if it solves your problem

e.g. you could say an engine with an ubershader has a hardcoded pipeline, but it's not very meaningful

but eh, it's not actually that meaningful of a distinction

cause the point of moving away from specifying heights on a set grid is probably a valuable one

[thick edge]

When I say hard coding I'm talking about stuff like having your renderer have an explicit step for drawing terrain or other things that intrinsically tie your renderer to the application

Rather than presenting a generic but more complicated interface that the game then uses as an intermediate implementation layer between the game and renderer

[inland flax]

i'm sort of just agreeing with this

LVSTRI — Today at 5:16 PM
Keep hardcoding stuff until it works is good tbh

[thick edge]

Ah yeah I see

That's basically what I do, I hard code and then start exposing control when it becomes necessary

[eternal dune]

I don't agree with it either generally, but in specific cases it's a good way of getting started, offloading the headaches of API plumbing and whatnot for later

For example, if you are a beginner and you are learning your way through the API

I completely misread your statement didn't I

lmfao

My bad

[inland flax]

:^)

i would bump this though and add some other stuff that might be fun

• add a player camera controller
• skyboxes
• terrain materials

if you did want to lean into terrain, you do have a whole bunch of people you can bug about it 🙂

[solid jackal]

imgui seems more trouble than its worth for this project since i dont really need it and i need to write a custom backend for it

there is a camera, not sure what you mean

[inland flax]

first person, walking on the terrain

[solid jackal]

because then i have to do physics

i mean i first envisioned this project as a photo realistic forest scene but since thats not a great idea

i guess i just want to try a ton of new things and experiment with what i have

[inland flax]

you do not have to do physics

but 👍 just ideas, pick what you like or ignore em all, no need to justify

[topaz cove]

skyboxes help the vibe so much

[verbal vector]

skybox /atmosphere for sure

[solid jackal]

perhaps

[verbal vector]

or terrain multitexturing

[solid jackal]

vulkan is good for different types of objects, right?

why not try billboards?

[verbal vector]

billboard what though. unless you make really distant trees

[solid jackal]

grass 🙂

[verbal vector]

oh, yeah. grass can be a beast

[solid jackal]

or a cool dinosaur idk

[topaz cove]

yo doom/daggerfall style sprite billboard dinosaur would add so much character lol

[solid jackal]

yeah to my windows xp forrest

is it official then?

am i going for the windows desktop aesthetic lol

wait

my brother in Christ this is my project

and I say yes

[thick edge]

Imgui is super easy to implement all things considered

It's just that Vulkan is hard in general

In GL it takes like 5 min to get it up and running

[solid jackal]

yeah

it would be pretty easy on vulkan but the backend doesn't support a vulkan feature that I'm using

Oh yeah baby

multisampling

(top image)

[verbal vector]

look at the little flowers!

[uncut tusk]

tre

[solid jackal]

blending moment

[solid jackal]

i think im multiplying my alpha when applying lighting

yeah its fixed

my leaves look a lot better now

[verbal vector]

oh they're certainly less distracting now

[solid jackal]

lol my girlfriend said this looks like the golf from wii sports

[solid jackal]

As is standard in vulkan, the next thing I want to add will take some time

remain patient

[uncut tusk]

What is the next thing that you want to add

[solid jackal]

shadows

im gonna need a new pipeline which is always fun

and ill need to work with multiple render passes which I’ve never done before

[uncut tusk]

adding a pipeline shouldn't be that bad

[solid jackal]

not really

youre right

[uncut tusk]

just should work a bit on getting a bit of structure in the project

otherwise it might take longer than it should

[solid jackal]

yeah

next time I might abstract pipelines

[uncut tusk]

Wasn't thinking in terms of abstracting pipelines but rather something that lets you structure the renderer as a whole easier

so you can focus on the important stuff

[solid jackal]

yeah

is this where rendergraphs come in?

they sound really complicated and confusing lol

[uncut tusk]

no need to do an entire rendergraph

For example I do something like a "flattened" rendergraph which I can later expand into one if I feel the need to

but to actually make use of a rendergraph you need enough "nodes" for it in the first place

[thick edge]

Pipeline creation is exactly the kind of thing you should be abstracting

Generally for something like Vulkan you want to follow a semantic compression technique

I have a PipelineCreateInfo struct that I defined that contains all the things I actually care to change in a pipeline, like the render pass, primitive topology, depth/stencil settings, shader info, etc., which just gets passed to a function that handles the rest

Anything I don't care to expose gets hard-coded in that function

[uncut tusk]

well yes but that falls into the "helper" category more for me than abstraction?

[thick edge]

Yeah I agree

I think abstraction is the wrong word to use here

What I'm against is generalization

Generic abstraction is usually a waste of time

Specialized abstraction/semantic compression/"helper functions" are critically useful

[uncut tusk]

what is the difference between specialized and generic abstraction here

[solid jackal]

Yeah im with demongod on this one

[thick edge]

Generic abstraction is like writing an abstraction layer that attempts to simplify the use of Vulkan without sacrificing generality, i.e. you could use it to build any Vulkan application

[solid jackal]

I don't think there's any need in making some advanced pipeline class if im only having like 6

[uncut tusk]

I mean when I say "don't abstract pipelines" I don't mean don't do helpers for it 😄

[thick edge]

Specialized abstraction is like "rather than copying and pasting my hard-coded pipeline creation, I'm going to dump it into a convenient function"

Generic abstraction is like trying to roll your own OpenGL using Vulkan

Specialized abstraction is like writing a game directly in Vulkan but hiding the implementation details behind an API firewall that keeps the mess out of your game code

[solid jackal]

that's exactly what I've been doing lol

main looks like this

I should probably keep my engine state in a struct next time 💀

[uncut tusk]

So I suppose you would also count my D3D12 abstraction more as a specialized abstraction? It's essentially just a list of virtual calls to "execute()" functions so I can write multiple demos using the same architecture without having to rewrite loading and buffer updating etc.

[thick edge]

In some sense yeah

I'd have to see it I guess

Generic abstraction usually looks like an API that you would provide in a library, a sign that you're doing that is if you find yourself doing 2x the work

E.g. if you do a bunch of work to wrap your Vulkan graphics in a convenient API, and then you use that API to write a renderer, you're basically writing twice

Whereas if you just write the renderer directly in raw Vulkan + some helper functions you're not wasting any of that time

[uncut tusk]

The only thing I've wrapped was the command buffer and that's purely out of convenience (so I don't have to dance around levels of indirection with my resources anymore)

other than that it's practically pure d3d12

[thick edge]

yeah sounds good to me then

I mean ultimately all that matters is whether it serves your purposes well and is efficient to use

[uncut tusk]

Yeah efficiency of use is what I care most about

and as time and need comes I enhance the API for more functionality, say something like "instead of just a list on a single queue I could allow for async compute processes"

[thick edge]

The problem with generic abstraction is that you end up fighting your own abstractions, like "hmm I designed my API around this assumption but now I need control over this hidden thing and it's becoming a pain in the ass," whereas with helper functions you're just like "ok I'll just expose this hard-coded thing as a parameter and the 'user' [me] will just set it manually"

[uncut tusk]

I do expose the raw D3D12 stuff if I need to which helps that but I've noticed as well in one scenario where it did bite me

[thick edge]

Well it only works if you can pull your graphics code into a black box

What I do is write my game to have essentially no conception of graphics except for physically-relevant data (e.g. meshes, needed for physics/raycasting), and everything else is just an ID

The renderer is content-aware (i.e. it's written bespoke for this game) and just handles raw Vulkan/GL + helper functions

That way none of the impl details leak out into the game, the renderer becomes part of the game itself (but stays isolated)

The difference is that if you wrote a generic graphics abstraction you might be able to replace the graphics API (e.g. swap GL for Vulkan) without touching the game renderer

Whereas in my case you have to rewrite the whole renderer

But it's a lot cleaner and doesn't restrict the API usage at all

a generic graphics abstraction designed around GL is basically just going to be like using GL if you rewrite the backend in Vulkan

Whereas in my case I can rewrite the whole renderer directly in Vulkan and it doesn't cause any architectural pain because the renderer is a black box

I've done this before btw I completely swapped the renderer and only touched 2 lines of code anywhere else in the codebase

And the whole time I have unrestricted access to the API like it's a greenfield project, not fighting any stale abstraction

[uncut tusk]

This is pretty much what I am trying to achieve with this "virtual execute() calls" - Having a renderer per-game and not one fixed for the engine. You decide what you need and extend if necessary.
What I did not want to do is having to redo allocators, binding models, resource management (to some degree), device/context creation etc. manually on a per "project" basis.

[uncut tusk]

Yeah hope I achieve the same there

[solid jackal]

i think i might get shadows in by tonight

i may have overestimated how hard it would be

[eternal dune]

Remember that y is flipped

[solid jackal]

Ah the path of tediousness

I can either add another descriptor to my global descriptor set or just push constants, but that would require a new pipe layout for the shadow pass

either way im tired

im gonna go play minecraft

[eternal dune]

Descriptor Sets are very annoying indeed

They're something you abstract very early or just use push descriptors

[uncut tusk]

Just use Bindless (tm)

[eternal dune]

BDA will save you from descriptor sets

[thick edge]

I didn't find them that bad for my simple purposes, you just have to group by update frequency

[uncut tusk]

Who needs pipeline layouts when all you have is a single pipe layout

[uncut tusk]

It isn't that bad once you got the boilerplate for them yeah

but first you need to write the boilerplate for them

[eternal dune]

The only descriptor sets I have are for storage images, samplers and textures

Because there's no IDA (Image Device Address)

[uncut tusk]

I don't have any descriptor set

[eternal dune]

OpenGL has Image Device Address btw, clear win for OpenGL

[uncut tusk]

But I also don't have BDA

[solid jackal]

my push constants are already bursting with data

im budgeting hard

the shadow pass will need 2 mat4s, while the render pass will need a mat4 and the texture index

and imo descriptor sets aren't the most annoying thing in vulkan

[eternal dune]

I recommend push descriptors if you don't want to deal with the tediousness of descriptor sets/pools

They're suboptimal, but extremely unlikely to make a difference in performance

In exchange you gain ease of use

[eternal dune]

How are shadows going @solid jackal

[solid jackal]

im still figuring out how to get my matrices to work lol

when i get back to my pc ill show yall the problem

[solid jackal]

im going to do something very lazy

[thick edge]

What's that

[solid jackal]

im putting the sampler to my shadow map at the very end of my texture array lol

[eternal dune]

Did you manage to fix the proj

[solid jackal]

yeah I think so

[eternal dune]

Let's see the shadow map

[solid jackal]

program on life support

im about to do the first test

[solid jackal]

shadowing isnt going well

[uncut tusk]

you're supposed to darken the areas in shadows

[solid jackal]

reaaaalllllyyyy?

[uncut tusk]

Yeah I know that's astonishing right?

[solid jackal]

alright so it looks like the transform data is correct

[inland flax]

i'm not gonna read all of #questions , is your sun's depth buffer reasonable?

https://cdn.discordapp.com/attachments/591343919598534681/1126222626495418418/image.png

nvm i see you got it

[solid jackal]

Vulkan Forest Scene

[thick edge]

Scope adjustment?

[uncut tusk]

Seems like it

[solid jackal]

its not gonna me massive in terms of size anymore

unless you guys think it should

[thick edge]

Lol just teasing

Do whatever makes sense

You're finally discovering what the real graphics programming is like

[solid jackal]

Nice

i think i want a skybox next

i dont think vulkan has cubemap functionality though

so i’ll need my own

currently im cleaning and refactoring some code

more specifically im putting most of my data in a big struct so i can spread flow out over functions more evenly

[eternal dune]

Yes it does

[solid jackal]

oh?

ill have to look into it then

[solid jackal]

Graphics::run() is now less than 100 lines

it hasn't been this way in so long

[solid jackal]

Alright, it's pushed

https://github.com/GrumStomper29/Vulkan-Forest-Scene

[thick edge]

Cubemaps are just a sampler type

[solid jackal]

I'm gonna be on vacation all of next week

[inland flax]

did you book your time off

[thick edge]

Yes please submit your request so we can approve it

[solid jackal]

no

[solid jackal]

i’m back

[thick edge]

Welcome back

[inland flax]

im gone

[verbal vector]

ready to see more 🌲 🌲🌲🌲

[solid jackal]

yeah

[solid jackal]

i completely forgot what i planned next

cubemaps?

i wish i could work more on this than i can

[solid jackal]

i cant find any good skyboxes

oh well

I kinda wanna make a huge forest scene in blender but due to .obj it will probably take several gigs to load lol

[thick edge]

In a game you wouldn't have a unique model for the trees anyway, you'd have 5-10 tree models that you'd instance

Maybe with some procedural scaling or coloration applied

[solid jackal]

yeah but idk if obj has instancing

[thick edge]

what? obj is a mesh transmission format

It has nothing to do with instancing

Instancing is what you develop in your engine

You need to develop your own asset pipeline around whatever you want to do in your application

[solid jackal]

right

im not going to explain what i meant because i’m tired

gn

[solid jackal]

I meant like

duplicate meshes

I think in obj if you wanted an identical mesh in different locations you would just have to duplicate the all of the vertices and indices

[eternal dune]

ye

[thick edge]

Different locations? Export each tree as its own obj file

And then instance them in your application

[eternal dune]

with obj you make your own scene representation yes

you have one obj per unique mesh

and instance them yourself

[solid jackal]

my normals are screwed up

i suspect its because they arent being rotated along with the model matrix

[verbal vector]

you need to apply a TBN matrix

also you need to decide to either transform obj to world or light to obj space. iirc obj to world can cause precision problems but it works for me

[solid jackal]

whats a tbn matrix?

I think

I want to finish this simple lighting, add rain, and ship it

[verbal vector]

tangent bitangent normal matrix

[solid jackal]

does glm have a function for this?

[verbal vector]

no look it up. it's a standard equation for applying normal mapping to geometry

[solid jackal]

ah I see

would this be done in the vertex shader?

ah

praise learnopengl

but I need 2 more vec3s per vertex 😦

[thick edge]

You don't need a TBN matrix if you're not doing normal mapping

Just rotate the normals by the model matrix

[solid jackal]

oh

ill try that lol

[solid jackal]

uh question

it didnt work

[thick edge]

Idk your transformation chain you kinda have to debug it yourself

[solid jackal]

is there any special way I should be doing this?

because rn I'm just multiplying it like its a position vector

[thick edge]

It's not a position though it's a vector

[eternal dune]

it's just normalize(inverse(transpose(mat3(transform))) * normal);

[thick edge]

Multiply it like (v, 0)

[solid jackal]

this is still genning the same results

now this matrix does have translations but shouldn't that not matter since we're ommitting the fourth row and column?

[thick edge]

Yes

Better check the values then

[solid jackal]

alr

heres some samples from the lab

now I may be wrong

[thick edge]

I meant you do it lol

[solid jackal]

well yeah im just tossing these here

[thick edge]

It's not really possible to debug someone else's numbers unfortunately

[solid jackal]

yeah

now I may not be an expert

[solid jackal]

but these don't seem normalized at a glance

and I think that may make sense since these are tree normals

and the trees in the scene are noticable darker than the ground

[thick edge]

Yes those are not normalized lol

[solid jackal]

then we have a problem

could this be because the matrix also has scale too?

ok its fixed

sunset 🙂

[solid jackal]

I need to fix shadows

[solid jackal]

not a fan of the ground texture

[verbal vector]

it's okay. gives me some late 90's vibes

[solid jackal]

I think i might be done with this project

Today i might clean up the code, and push it to the github

ill see if my models can fit on github too so yall can run it

[solid jackal]

It's been a wonderful vulkan forest scene

[verbal vector]

do you feel like you learned a lot?

[solid jackal]

Yes

practical and theory

[eternal dune]

I wish I could end a project

nice work fam

[solid jackal]

ah my .obj is over 100 mb

let me try and compress it

I probably can't

[eternal dune]

You could definitely come back to this when you gather the motivation, here's a few things you can add that are not trivial:

• Infinite world with world partitioning
• LODs

[solid jackal]

I can fit in the .blend file though LMAO

ill just include instructions on setting it up

Alright guys

all yours if you wish to try it

https://github.com/GrumStomper29/Vulkan-Forest-Scene

[thick edge]

What's next?

[solid jackal]

I'm making a gltf renderer in opengl just so I can learn how the both work

[thick edge]

I would just do the bare minimum gltf for now don't worry too much about the PBR stuff

[eternal dune]

OpenGL is great

[thick edge]

Even after learning Vulkan I've never really found a reason to justify using it

I just use the lessons it taught me to build more well organized renderers that could be swapped for a Vulkan impl later if necessary

GL is everything I've ever needed for my projects

[solid jackal]

yeah lol

[solid jackal]

im struggling to load a simple model

[thick edge]

Use a simpler model

You need something where you can inspect all the data in the file numerically

I've never used gltf so I can't really help beyond that

[uncut tusk]

rt/mesh shaders/some more advanced indirect stuff (ogl spv extension bas some bugs so subgroup stuff may or may not work), close to fully bindless

explicit async compute / copy as well

if performance is no issue and none of those things here are desired or required then yes, no real need to switch

[thick edge]

Yeah RT and mesh shaders would be some, I haven't needed them though

And if I did want to use RT for something I'd rather do it in software so that the performance isn't tied to having a certain kind of specialized GPU

[solid jackal]

I mean

hmm

should I close this?

since I guess im complete?

i might come back to it later but idk

[thick edge]

You can just leave it as is and come back to it later, you don't have to close the thread unless you want