Joe
983 posts

Joe
@Copper_irl
CSULB Physics and Applied Mathematics student, class of 2024(?) Tried for a CS degree for a decade, pivoted to physics. I really like computer rendering
Los Angeles, CA Katılım Eylül 2009
2.1K Takip Edilen217 Takipçiler

This one was fun! Didn't take too long either. Definitely not hours and pages of work. It's the negative of the laplace transform of cosine, evaluated from a to b. Feynman integration strikes again

Mathieu@miniapeur
Redondo Beach, CA 🇺🇸 English

youtu.be/Y6bWUfmJ0-4
My solution isn’t a clean formula, it’s code, so I’m a bit apprehensive to share it. I swear I’ve solved it before too, the same way, to render exact per pixel umbras back in high school. Should I prepare it nicely or just share what I have?

YouTube
English

@lisyarus Since this is a cpu rasterizer, heres a fun experiment if you can get shaders working again: Are shader quads ideal? I know this is how gpu hardware works, but I'm curious if there's any performance advantages to using differently sized pixel groups like 2x1, 3x3, 16x9, etc.
English

I think in my attempts of explaining things as thoroughly as possible I actually made them way more obscure 😅
Time to rewrite this section from scratch I guess
Nikita Lisitsa@lisyarus
I'm in the process of writing the quaternions article and oh boi I'm spending so much time explaining something instead of just saying "4-dimensional real associative algebra" in the hopes of making the article be much more accessible 😭
English

@AlexanderRKlotz @PolymerPolymath Njit is a bit picky but it is comparable to c in performance when it works. I’ve used cppyy before as an alternative but it’s way pickier. I think the right decorators help njit compile properly but I usually have to write c-like code for it to be happy
English

@PolymerPolymath I pasted my code into their example notebook and it gave me weird errors, so that's enough for me.
English

My MATLAB code computes the 2nd Vassiliev invariant for a 100-vertex trefoil knot in 0.026 seconds. My Python code does it in 2.89 seconds, 108x slower. Time to learn C?
Alex Klotz@AlexanderRKlotz
I have started giving python a serious use after spending my entire career using MATLAB and would like to report that I am, in fact, a dingus.
English

@BenSimsTech Oops, it’s gohugo.io. That’s what I get for manually typing it instead of copy pasting
English

@BenSimsTech No personal experience, but some devs I know use hugo.io because it lets them use markdown easily. There’s a ton of options here to generate static sites apparently. Hugo’s latex integration looks easy, so it’s my first pick when I get around to making one
English

@sp_monte_carlo Whoops! Been using that one as a variant of capital I for so long I completely forgot its capital phi. Full time job as Flux manager makes it very different to the angle triplets

Redondo Beach, CA 🇺🇸 English


For fun I did a radius inversion (r → 1/r) to watch the behavior as the ray diverges. It appears that the sdf implicitly includes the point at infinity, so maybe PGA is better for sdfs in general? I've really gotta get @EricLengyel's book on this, it looks fantastic
English

@BenSimsTech How do you feel about immediately aliasing these variables? For example with a derivative of smoothstep:
res = 6*timerElapsed()*(1 - timerElapsed())
imo the shorthand is easier to read in this case but i can def see where your coming from
t = timerElapsed();
res = 6*t*(1 - t);
English







