Pretty cool N-Body simulation that can do a bunch of stuff.

You can look around, the code is heavily commented. However this idea is bad, like really bad. This effectively turns a perfectly fine GPU into a glorified CPU.

There isn't any kind of forces computation optimization, ie we don't use the interactions symmetry

Might be usefull since I run on a single SM, CF fortran N-body to check if that damn race condition still mess around

Can add different kind of interactions

Euler and Leapfrog integrators are available. Might add RK4(5) but ain't for now.

Went with a simple integrator instead of doing object oriented with p1.interact(p2)

---

Env vars

#!/usr/bin/env bash
# ================================
# Simulation Environment Settings
# ================================

# --- Simulation settings ---
export SIM_NUM_SYSTEMS=24
export SIM_PARTICLES_PER_SYSTEM=2048
export SIM_TOTAL_STEPS=10000
export SIM_STEPS_BETWEEN_RECORDINGS=100
export SIM_DT=0.001

# --- Enum settings ---
export SIM_INTERACTION="gravity" # Gravity or RepulsiveForce or IdealGas
# export SIM_INTEGRATOR="verlet"   # Uncomment if get_env_integrator() is implemented

# --- Interaction parameters ---
export SIM_EPSILON=1.0
export SIM_SIGMA=1.0
export SIM_GRAVITY_G=1.0
export SIM_GRAVITY_SMOOTHING=0.01

# --- RNG settings ---
export SIM_RNG_SEED=42

---

Mixed Precision analysis

Gonna need to parametrize :

particle_system.h

position, velocity, acceleration should be added to the template => No issue with Vec2

boundaries.h

Need to pass particle precisions box_max and box_min most likely don't really matter, maybe check what comparison is best depending on Particle.position precision is? ie is float>double worse than double>double or something ........

interactions.h

RepulsiveForce

Need to pass particle precisions

Could be nice to add epsilon and sigma to see if precision here would impact anything (best bet is we could go wayyyyyyyy down, float8 or something even?) I'm sure there's some dumpster diving to do in this one.

IdealGas

Need to pass particle precisions

Not much to do

Gravity

Need to pass particle precisions

Try to parametrize G and smoothing_factor_sq

Might be easier to just change it in the function tbh

integrators.h

Need to pass particle precisions

EulerIntegrator

Should depend on particle precisions when I refactor it

Spoiler, I probably won't because Euler sucks

LeapfrogIntegrator

Need to pass particle precisions

Parametrize dt too? Seems actually duper important

Simulation.cu

Need to pass particle precisions

F my sanity

There's so much to do here

I need to get rid of the switch and make it cleaner

Simulation.cuh

Need to pass particle precisions

Update templated declarations ;(

main.c

Tough nut, will change during implementation