hyperstellar 0.1.81

GPU-accelerated physics simulation engine

Hyperstellar

Write math Equations in python

Buckle up, because this isn't just another preset physics engine. Hyperstellar gives you the mathematical language to define any dynamical system, then GPU accelerates it to thousands of frames per second. From orbital mechanics to fluid dynamics, if you can write the equation, you can simulate it.

Installation

pip install hyperstellar

Platform Note: The current release (0.1.x) supports Windows 10/11 (64-bit). Linux support is in development - Linux users can build from source using the project files.

Quick Start

import hyperstellar as se
import math

sim = se.Simulation(headless=False, enable_grid=False)
while not sim.are_all_shaders_ready(): # A one time GPU initialization (required before simulation)
    sim.update_shader_loading()
while sim.object_count() > 0: # Remove default object
    sim.remove_object(0)

G, M_star, M_planet, sep = 1.0, 50.0, 1.0, 3.0
v_orbit = math.sqrt(G * (M_star + M_planet) / sep)

# Create star and planet
star = sim.add_object(x=0, y=0, vy=M_planet*v_orbit/(M_star+M_planet),
                      mass=M_star, skin=se.SkinType.CIRCLE, size=0.8)
planet = sim.add_object(x=sep, y=0, vy=-M_star*v_orbit/(M_star+M_planet),
                        mass=M_planet, skin=se.SkinType.CIRCLE, size=0.25)

# Gravitational force equations (ax, ay, angular, r, g, b, a)
sim.set_equation(star,
    f"{G}*{M_planet}*(p[1].x-x)/((p[1].x-x)^2+(p[1].y-y)^2)^1.5," # Newtonian gravity equation using object reference p[index]
    f"{G}*{M_planet}*(p[1].y-y)/((p[1].x-x)^2+(p[1].y-y)^2)^1.5,"
    "0, 1.0, 0.9, 0.3, 1.0"  # Yellow
)
sim.set_equation(planet,
    f"{G}*{M_star}*(p[0].x-x)/((p[0].x-x)^2+(p[0].y-y)^2)^1.5,"
    f"{G}*{M_star}*(p[0].y-y)/((p[0].x-x)^2+(p[0].y-y)^2)^1.5,"
    "0, 0.3, 0.6, 1.0, 1.0"  # Blue
)



while not sim.should_close():
    sim.update(0.016)
    sim.render()

See the full version in examples/orbit.py

What you can actually do with this

• Write your own equations (real or complex) and run them directly on the GPU.
• Reference other objects inside equations — positions, velocities, even colors. p[i].value
• Reference other objects inside equations — positions, velocities, even colors. "p[i].x, p[i].y, p[i].vx, etc."
• Define derivatives and run them in the GPU. D(expression, variable, order)
• Most Math functions are available in DSL. Full support for trig, exponentials, powers, statistics, and complex numbers
• Run the engine with a window, or completely headless if you just want numbers.
• Batch operations for crunching numbers and cpu to gpu uploads in mass.
• Use color as part of the simulation state, not just rendering.
• Lock things together with constraints instead of hardcoding relationships. -experimental
• Run thousands of objects or many simulations at once without touching the CPU.

Examples

• Two-Body Orbit - Newtonian gravity simulation
• Pendulum - Spring-based harmonic motion
• Boids - Emergent flocking behavior with obstacle avoidance
• MCMC Sampling - Metropolis-Hastings algorithm on GPU

Documentation

Coming soon - for now see the examples folder and the code documentation in bindings.cpp.

---

Built to make physics equations visual and interactive.