Voxel Editor update: dim sun fix, faster GPU path tracing with atmosphere volumetric shadows, and Minecraft ice blocks. https://www.enkisoftware.com/t/6272880503422976?count=10&start=61#6239597409337344
Download Avoyd Beta 0.28.0.1063: https://www.enkisoftware.com/products#downloads
#Avoyd #Voxel #VoxelEditor #VoxelArt #3D #Rendering #PathTracing #MinecraftRender #Minecraft #IndieDev #GameDev #ScreenshotSaturday
After ~6 years of development time today Keen Games finally released Enshrouded.
Congratulations to our entire team and absolute great work from everyone!
@dougbinks Our policy isn't long, but could also be trimmed down if brevity is the goal:
Relentless AI scrapers pushed DNA Lounge's web hosting bill from $800 to $3,500/month.
That is a $32,000/year cost shift caused entirely by companies training models on their content without permission. AI companies are externalizing their massive training costs onto independent web operators, effectively taxing the independent web to build their proprietary models. 📉
@dougbinks mine is this, it basically defines a 'no-llm-zone' for the actual library code, but allows LLM code gen for (some) other areas:
@dougbinks A simple one from Alacritty https://github.com/alacritty/alacritty?tab=contributing-ov-file#llmai-contributions
@dougbinks SDL has a good one https://github.com/libsdl-org/SDL/blob/main/AGENTS.md
@dougbinks
this is what I use in dhewm3: https://github.com/dhewm/dhewm3#contributing
https://github.com/dhewm/dhewm3/blob/master/.github/pull_request_template.md?plain=1
no idea if that's "good", seems to work so far
RE: https://mastodon.gamedev.place/@dougbinks/117438273462080269
I managed another obvious small perf improvement to this code - do not launch a shadow ray-cast if the contribution from path weight and in-scattering would be negligible. It's only about 1.04x perf improvement on my tests but it's very consistent and pure ALU to reduce ALU + memory traffic so likely a good future bet too.
Does anyone have a good example of a simple and succinct No Gen AI / LLM clause I can add to my open source repos? I'm also going to ask PR authors to declare they did not use Gen AI / LLMs (or if they did I will close the PR).
The indie hit RimWorld just made a MASSIVE change... they dumped the Unity game engine... and saw a doubling of performance.
Changing game engines 8 years after you publish is... brave!
https://gamefromscratch.com/rimworld-dump-unity-game-engine/ #gamedev #indiedev #unity3d
I'm considering moving to Slang to benefit from module support amongst other things, but for now creaky old GLSL is manageable https://shader-slang.org/
The infrastructure around GPU compute programming is still a mess in my view. It's easy to see why CUDA and NVIDIA dominate certain markets.
An easily measurable part of the shader compilation is how long it takes to compile and optimize the GLSL code to SPIR-V, which we only do on dev machines or a final build. This has gone from 200s to 45s, so a 4x improvement.
Worth noting is that prior to this change my shader had hit an ID Overflow issue in spirv-opt, resolved by using the same optimization with compact-ids found by @Blender contributor Ray Molenkamp https://projects.blender.org/blender/blender/issues/144614
The new shaders can use default `-O` optimization.
I've been asked by several clients over the past year to help them find senior software engineering talent. You need to play down the "AI" rhetoric if you want to attract people who genuinely care about the work and who really know what they're doing.
Advertising that you intend to install them into some kind of "reverse centaur" role will pretty much guarantee they won't be applying.
Call me a masochist, but I'm returning to this issue, as I feel it has been poorly understood. Please read the whole thread before responding.
First, I hope we can agree that the Israeli government is committing genocide in Gaza and horrific crimes in the West Bank and the wider region.
the mass production of part of software development has largely been abstracted, automated, standardized away: distribution of boundlessly replicable bit-perfect copies of software over the internet. that gives us the illusion that more automation is needed, since the remaining work seems manual by comparison. but you can’t and shouldn’t automate it; it’s design, feedback loops, thinking, creating, problem solving; all very irregular, reactive, and unpredictable types of work
Drehmal map from: https://www.drehmal.net/
Megakernels Considered Harmful: Wavefront Path Tracing on GPUs by Laine et al.: https://research.nvidia.com/sites/default/files/pubs/2013-07_Megakernels-Considered-Harmful/laine2013hpg_paper.pdf
PBR book online chapter discussing mapping path tracing to GPUs and Wavefront path tracing: https://www.pbr-book.org/4ed/Wavefront_Rendering_on_GPUs/Mapping_Path_Tracing_to_the_GPU
This optimization is not yet out in Avoyd Voxel Editor but it should be in the beta soon.
https://avoyd.com/
Finally got around to moving the atmospheric shadow loop in Avoyd's GPU path tracer to a wavefront approach like the rest of the code instead of a loop. I was mainly doing this to improve the insane shader compilation times, but it also improves hardware utilization. Measuring compilation perf is tricky as the driver caches pipelines, but I've measured a consistent 2x render perf increase on top of a recent 3x perf increase.
#Avoyd #Voxel #Minecraft #VoxelEditor #Pathtracing #Raytracing
Data:
- LLMs are killing the planet
- LLMs are raising utility prices
- LLMs are raising computer prices
- LLMs give you teh dums
- LLMs show negligible increases in net productivity
- LLM costs are going up
- LLMs run on stolen content
- LLM companies keep breaking laws and saying, "Oopsie!"
Open Source maintainers for some reason:
- We are debating our AI policy, but it's a tough decision. *whines*
Dudes. WTF is wrong with you?