OwenTheProgrammer
@owentheprogrammer.bsky.social
Game dev / Graphics dev. All I know is math now.
id put an unpopulated resistor on ITERM for when you want to change the current from 10% threshold, also 10u decoupling capacitors are great, but they're better with ones that act quicker (10u + 100n type beat) although the datasheet has 1u on input, 4u7 on output.. so good jab
January 22, 2026 at 12:28 AM
id put an unpopulated resistor on ITERM for when you want to change the current from 10% threshold, also 10u decoupling capacitors are great, but they're better with ones that act quicker (10u + 100n type beat) although the datasheet has 1u on input, 4u7 on output.. so good jab
Ever wanted to store a quaternion / versor in 3 floats instead of four? there is actually a mathematically 1:1 (in practice e05 error terms for FP16, e09 for F32) representation using lie algebra.
Heres all the code to encode / decode that. Versor -> rotation axis with length of the angle.
Heres all the code to encode / decode that. Versor -> rotation axis with length of the angle.
January 11, 2026 at 6:57 AM
Ever wanted to store a quaternion / versor in 3 floats instead of four? there is actually a mathematically 1:1 (in practice e05 error terms for FP16, e09 for F32) representation using lie algebra.
Heres all the code to encode / decode that. Versor -> rotation axis with length of the angle.
Heres all the code to encode / decode that. Versor -> rotation axis with length of the angle.
P = MAN lol
December 4, 2025 at 4:19 AM
P = MAN lol
Probably not you, I couldn’t give a single fuck about 99% of my feed at the moment
November 22, 2025 at 7:39 AM
Probably not you, I couldn’t give a single fuck about 99% of my feed at the moment
I knew it would be directly equivalent, but im still surprised its directly equivalent lol
This actually might make things for latrix ever so slightly faster / easier to work with, although it does come with the prerequisite of complex analysis, since I cant preprocessor test for i^2=-1 it seems..
This actually might make things for latrix ever so slightly faster / easier to work with, although it does come with the prerequisite of complex analysis, since I cant preprocessor test for i^2=-1 it seems..
November 21, 2025 at 8:04 PM
I knew it would be directly equivalent, but im still surprised its directly equivalent lol
This actually might make things for latrix ever so slightly faster / easier to work with, although it does come with the prerequisite of complex analysis, since I cant preprocessor test for i^2=-1 it seems..
This actually might make things for latrix ever so slightly faster / easier to work with, although it does come with the prerequisite of complex analysis, since I cant preprocessor test for i^2=-1 it seems..
eeeeexcuse me?
November 14, 2025 at 5:41 AM
eeeeexcuse me?
@freya.bsky.social biblically accurate Unity gimbal
November 9, 2025 at 9:25 AM
@freya.bsky.social biblically accurate Unity gimbal
and after a quick verification, it seems to hold true
October 31, 2025 at 1:04 AM
and after a quick verification, it seems to hold true
I've been chugging away at the 3D one all day, but in the 2D solution, I may have found a property that makes things more elegant.
The first image is the actual integral (ln -> arcsinh), which is meant to handle one of the four triangle sweeps. The second image is the splitting of the unit square
The first image is the actual integral (ln -> arcsinh), which is meant to handle one of the four triangle sweeps. The second image is the splitting of the unit square
October 31, 2025 at 12:51 AM
I've been chugging away at the 3D one all day, but in the 2D solution, I may have found a property that makes things more elegant.
The first image is the actual integral (ln -> arcsinh), which is meant to handle one of the four triangle sweeps. The second image is the splitting of the unit square
The first image is the actual integral (ln -> arcsinh), which is meant to handle one of the four triangle sweeps. The second image is the splitting of the unit square
4 bits per sample, 2x2 block segments, 32 samples per colour channel, on a 512x512 CRT
October 9, 2025 at 7:44 AM
4 bits per sample, 2x2 block segments, 32 samples per colour channel, on a 512x512 CRT
wait yeah holy shit ive reinvented xray reconstruction like to a T lmao en.wikipedia.org/wiki/Tomogra...
October 7, 2025 at 5:56 AM
wait yeah holy shit ive reinvented xray reconstruction like to a T lmao en.wikipedia.org/wiki/Tomogra...
normal latrix on the left, the texture sampled one on the right
October 7, 2025 at 5:49 AM
normal latrix on the left, the texture sampled one on the right
I fixed the weird fringing from before 512x512 render texture, genuinely insane how it got to this point wtf
October 7, 2025 at 5:33 AM
I fixed the weird fringing from before 512x512 render texture, genuinely insane how it got to this point wtf
That normal reconstruction methods pretty solid (doing the normal reconstruction from the point of view of the laser) but its slightly weighted incorrectly, since it solves the direction vector to the tangent plane, not the planes normal itself. Ill try to fix that at some point
October 5, 2025 at 11:41 PM
That normal reconstruction methods pretty solid (doing the normal reconstruction from the point of view of the laser) but its slightly weighted incorrectly, since it solves the direction vector to the tangent plane, not the planes normal itself. Ill try to fix that at some point
Fresnel goes a long way
October 4, 2025 at 3:26 AM
Fresnel goes a long way
we'll also ignore the fact that the target for this math was.. not screenspace normals..
September 27, 2025 at 9:05 AM
we'll also ignore the fact that the target for this math was.. not screenspace normals..
It even does better on the edge of the screen! I could definitely do better for occupancy and register count
September 27, 2025 at 9:02 AM
It even does better on the edge of the screen! I could definitely do better for occupancy and register count
Cant be fucked to type this shit all out again, but ive made yet another "Screenspace normals from depth texture" method, thats actually somewhat stable
September 27, 2025 at 8:11 AM
Cant be fucked to type this shit all out again, but ive made yet another "Screenspace normals from depth texture" method, thats actually somewhat stable
this is beat saber. :3
September 20, 2025 at 12:42 AM
this is beat saber. :3
fun fact, you can do transparency blending with editor script stuff
September 12, 2025 at 2:46 AM
fun fact, you can do transparency blending with editor script stuff
it dooo move
September 9, 2025 at 1:02 AM
it dooo move
here's an artboard of my life in the shader compiler
September 7, 2025 at 12:35 AM
here's an artboard of my life in the shader compiler
today I made the CNE crew uncomfortable and confused by staring directly into an LED par light for five minutes straight. This also marks the first time ive ever seen any actual stage lighting irl 😂
September 1, 2025 at 2:28 AM
today I made the CNE crew uncomfortable and confused by staring directly into an LED par light for five minutes straight. This also marks the first time ive ever seen any actual stage lighting irl 😂
nahhhh there aint no way it look this good wtf
August 30, 2025 at 5:42 AM
nahhhh there aint no way it look this good wtf