Nikita Lisitsa
@lisyarus.bsky.social
He/him I teach C++ & computer graphics and make videogames Working on a medieval village building game: https://youtube.com/playlist?list=PLSGI94QoFYJwGaieAkqw5_qfoupdppxHN&cbrd=1 Check out my cozy road building traffic sim: https://t.ly/FfOwR
Almost finished a huge refactor to support bundling several source files into a single executable blob, carefully merging the respective code, data, and relocation buffers, while potentially allowing for parallel compilation. Important for future modules & executable (AOT-compiled) binaries support.
Love this project and writeup! A color space to cover skin colors. toneyalexander.github.io/inclusive-co... (can also be useful for #procgen)
What Colors Are We? Constructing A Color Space For Skin Tones
Simple equations and a color picker to inclusively select simplified skin tones in a custom color space for digital art or game development.
toneyalexander.github.io
I really need to add thousands of tests for this compiler, including testing the C FFI and ABI conformance and etc etc, but that's 1) a ton of work and 2) feels like committing to another not-just-for-fun project, and tbh I'm not sure if that's what I actually want rn
Nikita Lisitsa@lisyarus.bsky.social · 3d agoOhhhhh god I fixed it! The issue was that when comparing two registers I always compare them as 64-bit, sign/zero extending before that if needed. For unsigned 32-bit values I don't zero-extend as writing to low 32 bits of a register automatically clears the top 32 bits...
Aaand now a better image! 256x256, 8k samples, p=0.25 russian roulette (avg path depth=4), ran for 5m30s in 16 threads on a 3.6GHz i9-9900K, written in my own JIT-compiled programming language that now compiles to both aarch64 and x86_64!
Ohhhhh god I fixed it! The issue was that when comparing two registers I always compare them as 64-bit, sign/zero extending before that if needed. For unsigned 32-bit values I don't zero-extend as writing to low 32 bits of a register automatically clears the top 32 bits...
So my raytracer + compiler bug is unbelievably subtle. The image should be all-white but for some reason it isn't. Removing some lines (typically dead code - unused values / uncreachable instructions) removes the bug. I've spent way too many hours in gdb by now...
I've posted my first MinerMancers dev log! Finally! 😄 Do let me know what you'd like to hear about next, if you have a preference. If a particular subject gets more requests I'll be sure to prioritise it. Hope you enjoy! youtu.be/-wdz4vMpIzA
MinerMancers DevLog 1
YouTube video by Steve Streeting
youtu.be
Oh no: the bug is due to the missing u64 -> f32 conversion on x86_64 I tweeted about earlier. It's used in the RNG (roughly float(rand()) / UINT64_MAX)
Nikita Lisitsa@lisyarus.bsky.social · 5d agoTurned out I messed up the implementation of conditional jumps in my language's x86_64 compiler. Now the raytracer works properly! Meaning it runs in several threads, reports progress, and generates an image. The image is a bit off, though, but that's no big deal 😅
Turned out I messed up the implementation of conditional jumps in my language's x86_64 compiler. Now the raytracer works properly! Meaning it runs in several threads, reports progress, and generates an image. The image is a bit off, though, but that's no big deal 😅
Toy programming language compiler x86_64 backend status: implemented (almost) everything, the raytracer test program successfully compiles and runs! ...but never finishes and never prints anything and never creates the threads as it's supposed to, not to mention the output file 😅
New video: Supercharging my procedural mountain forest Two years of landscape progress in one devlog! www.youtube.com/watch?v=ACjl... #GameDev #IndieDev #ProcGen
TIL there's no unsigned 64-bit -> floating-point 64-bit conversion in x86_64, so you have to build it up from other instructions. Here's the codegen by gcc-15.2 with -O3, it uses different code paths based on the highest bit of the input:
The x86_64 compiler backend for my pet language is about 75% done - what's left are function calls, pointer load/store, struct function arguments, and some clean-up. Keep in mind that it's not even close to an optimizing compiler (yet), so the codegen is rather pathetic
I've started writing my language's compiler for x86_64 architecture (only supported aarch64 before) and oh god the induction encoding is such a shitshow
Currently the parser doesn't allow newlines before the opening curly brace {, which forces a certain formatting style. Not sure if I want to keep that or allow a different style, though...
I decided to ditch indentation-based scoping in favor of explicit curly braces. It's a bit more verbose, but to me it feels more readable. Maybe it's just a habit? Also the parser is 50% simpler now 😅
Been toying with the idea of rewriting my engine's debug renderer using SDFs supporting arbitrary affine transformations, and strokes (i.e. drawing the object boundary with some width) turned out to be much harder than I expected 😅
My friend an I took part in Kenney Jam 2026! Plan was to make a platformer with levels composed of musical notes. Due to a messy schedule we've only managed to publish a half-baked prototype, but at least we nailed the movement animations 😅 itch.io/jam/kenney-j... #indiedev #indiegames #gamedev
Stupid thing from my childhood: after a maths olympiad (which had 19 participants) 4 of us we were sitting in some room. One of the organizers came in and said "there should be more than 4 times more of you!". I replied "yeah but less than 5 times more". He said "fair enough" and left.
thicc husband and father@lukeisamazing.bsky.social · 2w agoonce again thinking about the best Lord of the Rings meme
- Yep...that fluffy white thing over there in the sky...definitely a cloud. I'll call it Jack. (yes I know the video is most probably about different cloud types)
Youtube recommendations are just...something Thanks mate I think I know how to *checks notes* connect dots with arrows
