clcell 1.0.1

OpenCL-accelerated cellular automata simulator

clcell

clcell is an OpenCL-accelerated cellular automata simulator for Python 3.

Features

• OpenCL-based hardware acceleration
• Custom rulesets via clcell.RuleSet
• Parallel simulations via clcell.CASimulator.batch_simulate

Limitations

• Only binary cell states
• No support for infinite grids
• Grid boundary cells must be unpopulated

Installation

Regardless of the installation method you choose, you will need OpenCL drivers for your hardware.

Using pip

$ pip install --user clcell

Building from Source

Requirements:

• GNU Make
• Futhark
• setuptools

Clone this repository:

$ git clone https://github.com/Foxbud/clcell.git

Enter the project directory:

$ cd clcell

Build and install this package:

$ make install

Usage

import numpy as np
import clcell

# Instantiate a simulator using Conway's Game of Life as the ruleset.
sim = clcell.CASimulator(clcell.LIFE)

# Create a randomized game state to use as a seed.
seed_state = np.random.randint(0, 2, (1023, 1023), dtype=np.int8)
# Pad state with zeros (required for now).
seed_state = np.pad(seed_state, 1, constant_values=0)

# Simulate 10,000 generations based on that seed.
final_state = sim.simulate(10000, seed_state)

# Create a batch of 1,000 randomized, padded game states to use as seeds.
seed_states = np.array([
  np.pad(
    np.random.randint(0, 2, (127, 127), dtype=np.int8),
    1,
    constant_values=0
  )
  for num
  in range(1000)
])

# Simulate 1,000 generations based on each of those seeds.
final_states = sim.batch_simulate(1000, seed_states)

Changelog

v1.0.1

• Simplified how the device code checks for boundary cells.
• Refactored device code to use grids of 8-bit cells as opposed to 32-bit.
• Fixed incorrectly placed size annotations in one device function.

v1.0.0

• Initial release.