Charlie Callahan

Game engine development, C++, lua, PhD Mechanical/Optical engineering, laser sensing

[Charlie Callahan @ccallac7]

By rewriting 4x4 matrix multiplication using SIMD I was able to increase the framerate of my game engine (written in C++) by 10-20% since its used so often

[Charlie Callahan @ccallac7]

@timsoret I know I would be addicted immediately lol need to stick to productive activities

[Charlie Callahan @ccallac7]

I’ve never played Factorio

But I know better than to even try it

[Angus (dirtman) @dirtman]

Wondering if I should put this in my pitch deck

[Charlie Callahan @ccallac7]

@jbaker_graphics Surface attenuation seems like a really sensible approximation.

Excited to see how this turns out! Very cool project

[Charlie Callahan @ccallac7]

@jbaker_graphics Super cool! i dont think ive ever seen someone actually simulate it like this (just pictoral representations in textbooks etc). Are you simulating absorption inside the material or just dispersion?

[Charlie Callahan @ccallac7]

@FilippoTarpini Which one is which? And how do you implement the one on the right. Looks alot better

[Charlie Callahan @ccallac7]

Thank god for imgui

[Charlie Callahan @ccallac7]

@yacineMTB Develop the algorithm in simulation, then build it IRL with a linear actuator and a 5x inverted pendulum

[Charlie Callahan @ccallac7]

@OlexGameDev Interesting. A pool allocator is just a more efficient way of doing that. It has O(1) complexity vs O(N) complexity for the linear search technique. But for a small # of entities it probably barely makes a difference

[Charlie Callahan @ccallac7]

Every so often Il go through my codebase and replace instances of new[] or std::vector with fixed size arrays that are statically allocated at startup

Removing heap allocations makes your game FPS more stable - smoother

[Charlie Callahan @ccallac7]

@OlexGameDev Looks pretty efficient. if you want to “deallocate” and reuse data chunks, you should check out a pool allocator (this is a perfect application for that). Its super simple to implement too, only a few lines of code

[Charlie Callahan @ccallac7]

@haydendevs Get yourself a Rigol

[Charlie Callahan @ccallac7]

@jere_codes I like the left one better, but it does have some minor visual issues on the silhouette

[Charlie Callahan @ccallac7]

@MegaPunkGames I usually start by just using std vector, then once i understand how its being used in practice. I will refactor it to use a fixed buffer size

[Charlie Callahan @ccallac7]

@OlexGameDev Im using Entt for the ECS but for components that are attached to large pieces of data (several MB), il usually just store a pointer to the data payload rather than try to store it in the component struct

[Charlie Callahan @ccallac7]

@OlexGameDev There must be a way to do it… what does the data structure look like?

[Charlie Callahan @ccallac7]

@OlexGameDev True, and most things are actually limited in how large they can be anyways. Makes no sense to take the performance hit just to avoid having an arbitrary limit

[Charlie Callahan @ccallac7]

@vareck not sure about what its called, but heres the shader, pretty simple

[Charlie Callahan @ccallac7]

Added a mip based bloom implementation

Does the blurring progressively over desampled buffers (2^-mip) to allow bloom to spread over the whole image cheaply

Scales as O(log(kernelSize)) and its super efficient due to z ordered texture reads

[Charlie Callahan @ccallac7]

@marcsh It works really well