So I was watching some shader-related video on yt where they explained that the heavy hitter on performance is sampling and not calculations.
So I decided to redo some stuff and holy shit I can render at full res with almost 200 fps with MORE detail.
So yeah what I did was basically rework how I generate and sample my noise which made it run so much better
(Also still a WIP the repetition will be dealt with and the values will be tuned for better details)
I added a setting so that instead of each step being the same distance you can chose how many steps it’s going to take and then it bases it on the distance betweeen the entrance and the exit to of the bounding box.
I also experimented with settings trying to make it look good as clouds and yeah its starting to look really good.
And lastly I added blur since the samples are kinda noisy
So I added bounding boxes like I have been wanting to and it was suprisingly easy using the slab method, I also added shadows which gives the coulds so much more depth
So for the bounding boxes I use the slab method which return the entry and exit position of a ray for a grid aligned bounding box for very cheap. It is also easy to see if it doesn’t intersect because the exit position is closer than the entrance position so you can return early.
For the shadows I do another lower quality raymarch for each step in the main raymarch towards the sun to see how much cloud is in the way.
I am also having some problems with the Henyey Greenstein function returning abnormally huge numbers so that is turned off rn while I investigate further.
I was finally able to port the volumetrics to the compute shader which allows it to run at 1/2 or 1/4 res improving the performance greatly. I also tried making a cloud layer which is what you can see in the image.
It still look kinda bad but trust it will become better with time. It’s also running at a stable 200 fps with 1000 max distance which is great!
Next I want to add proper bounding Boxes since this is just skipping the step if its outside the y range of 60 - 70 which isn’t very efficient.
I also want to add internal shadows so for each step in the raycast it will send out another raycast to see how much light it gets, hopefully that can make it less flat.
#Devlog 5.5 - I hate unity
I just wanted to post a devlog since I haven’t done so in a long time but also I haven’t added a complete feature so that’s why I did 5.5 instead of 6.
Basically I have spent a couple of hours trying to get the worldpos in a compute shader since it’s a whole different process compared to the normal shaders.
It’s honestly insane that something that takes like 3 lines of code in one shader takes like hundreds distributed through multiple scripts in another.
Anyways now that I have the worldpos in the compute shader the rest should be a breeze!
(Image showing the texture generated by the compute shader in which red is the decimal part of x, green is the same but for y, and green for z)
#Devlog 5 - Major fog improvements
Added:
So since last devlog I have made the fog pickup lights and shadows from the main light, it’s very easy to do since unity provides the values for me.
I also added perlin noise to the noise generator along with the worley, I use the perlin as the base for the fog and the worley to make the edges more detailed with remapping.
I have also made some smaller optimizations like the raymarching stopping if the transmittance (the value which decided how much fog there is) gets so high that any additional fog wouldn’t make a noticeable difference.
Side note: My friend @Lazyllama is going to join me and he will be making more of the game mechanics so that I can focus on this.
So basically during some time I had no motivation at all to work on the game, I think it’s mostly because I spent this whole spring making a game whh I wasn’t really happy with.
Anyways I decided to force myself into making it and while adding some simple fog based on distance I decided that it didn’t look right and went down a rabbit hole of volumetric fog (procrastination).
Anyways I realized that while I was procrastinating I was genuinely interested in this and other things that I guess would fall into the category “Graphics programming”.
Anyways what I wanted to say was that the game will most likely not be a fishing game, maybe there will be some kind of game mechanics added but I’m going to focus on learning more of what I find interesting.
Also During the time between this and last devlog I have been working on volumetric fog and generating noise through fullscreen shaders and compute shaders. Right now I use worley noise for the fog although it still look very bad since it is IN DEVELOPMENT. Anyways thanks for reading and I hope the next devlog isn’t for 8h and also has some better images! (Yes this is volumetric I know it doesn’t look like what you would associate with volumetric rendering since it isn’t affected by lighting and stuff and also looks kinda 3d I don’t know what’s going on with that)
I finally added some movement as I felt it was important to see how the ocean looks from the players perspective.
I had, and still have, some issues with it which will be fixed when I figure out what is causing them.
I will try to implement a LOD system for the ocean because I think the huge ocean is what is making the game stutter.
Also I’m not sure if you can upload videos so this will be a test for that.
I did some major changes to how the water looks, first of all I increased the amount of waves to make it less repetitive. I now have 40 waves with random direction and magnitude.
I also changed the water to have a more low polygon look using per face normals. I made this change because I am going to make all the assets myself and as a beginner in 3d modelling I should stick to simple models, and my game should follow that theme.
I want to add a system that allows me to make waves randomly as I do have to make like 4 numbers for each wave that all work together.
Also adding some actual gameplay would be nice.
I decided to start working on my project today since I had literally nothing else to do. So i booted up unity and got started with making some terrain which turned out to be not terrible.
After that it was time for water, since my game is a fishing game I decided that water would be pretty important, and so I would spend a lot of time making sure it looks good.
I did some research on how other games handle water, and after deciding I in fact did not want to struggle with FFTs i decided to use the sum of some Gerstner waves (a.k.a trochoidal waves).
I booted up desmos since it’s great for solving the math part before moving on to the coding part, and after getting something that worked I started working on making it to a shader.
( Image 1: desmos waves )
After playing around with the shadergraph to make sure I understood how it works I got started making the actual waves. I had to make a custom function in HLSL both for vertex displacement and for getting the normal values.
After hooking it all up I had some actual waves. Now of course this didn’t go as smoothly as described, as most of my time went into debugging and trying to understand how shadergraphs work.
( Image 2: waves in unity!!! 🥹🥹🥹)
So after this I want to make the water look like actual water and maybe take a look at making the waves less repetitive, although it is pretty hard to see from the players perspective. I will make sure to come back to this when I have a better idea of what it will all look like.
Anyways I’ll see you later and I hope you have as much fun making stuff as I do! ✨
