About

A C++11 implementation of simplex noise on the CPU, created for the purposes of learning and expanding.

Download

The source for this library comes in 3 separate files: a header, a implementation file for templated functions, and an implementation file for non-templated functions.

SimplexNoise.hpp – The primary header file
SimplexNoise.t.hpp – The template implementation source
SimplexNoise.cpp – Non-templated implementation source

Features

Templating
This implementation is highly templated, meaning that both floats and doubles can be used without rewriting code.
Octaves
Simple octave code is included. This can be used as-is, or be modified to have more complex options and outputs.
1, 2, 3, and 4 dimensional noise
Hardcoded functions for 1 through 4 dimensional noise, the more commonly used noise functions that provide acceptable visual quality.
Generic n-dimensional noise
A function for any n-dimensional noise. While the gradients generated might not be as uniform as the fixed permutation table implementations, it still provides acceptable results for basic testing.
First order partial derivatives
All noise functions for dimensions 1 through 4 provide overloaded functions for returning the partial derivatives of the function at the given input point.
Custom permutation table
A custom permutation table can be input into a noise generating instance. This functions like a random number generator seed and changes the order the gradients will appear in.

Documentation

The following functions are available in the SimplexNoise API. Note that many of the noise functions are very similar, usually just adding an additional point component to the argument list, additional derivative output, or an octaves parameter.

The extended descriptions are minimized by default for ease of locating functions by name. Click the +/− or the function name to expand/shrink the description. (expand all | shrink all)

SimplexNoise class

The SimplexNoise class is the primary noise generation class. Once instantiated, it can have its noise functions called to generate noise.

Executables

Source codes and binarys for 3 executables are included in this project. They are used for testing various functions and brute-force solving noise function minimum/maximums.

test.exe

test.exe is an application to test the 1 through 4 dimensional noise functions, writing the output to a .pgm image. It has the following command line syntax:

test.exe filename width height scale dimensions octaves derivative

filename – the output filename of the .pgm file
width – the width of the output file, in pixels
height – the height of the output file, in pixels
scale – the scaling factor for the noise function coordinates
dimensions – the dimensions of the function to test.
Can be 1, 2, 3, or 4.
octaves – number of octaves to use
derivative – the number of the derivative to output.
0 outputs the noise function itself.
1 outputs the first partial derivative.
etc...

demo.exe

demo.exe is an application to test the generic noise_n function, writing the output to a .pgm image. It has the following command line syntax:

demo.exe filename mode norm_mode x y width height scale max_radius power dimensions octaves [x y z w ...]

filename – the output filename of the .pgm file.
If set to "-", it will output to stdout.
mode – pgm output mode; either binary or ascii.
norm_mode – normalization mode of the noise values.
centered will normalize around 0.
full will normalize to the range [0, 1] regardless of where the center is.
The differences between these two will often be marginal.
width – the width of the output file, in pixels
height – the height of the output file, in pixels
scale – the scaling factor for the noise function coordinates
max_radius – the maximum distance squared from a simplex point that a contribution can be non-zero
power – the exponent of the noise function's point contributions
dimensions – the dimensions of the function to test.
octaves – number of octaves to use
[x, y, z, w, ...] – Optional starting values for the first coordinate.
If omitted, they default to 0.

solver.exe

solver.exe is an application to brute-force compute the maximum/minimum values of a noise function. It does this by selecting permutations of gradients and iterating over the points of a simplex, computing the noise function at each point.

This takes very long for the 4d case, but the minimum/maximum values will converge to a good approximation in a reasonable amount of time. Therefore, the execution can be interrupted to get a good approximate value.

solver.exe show_every threads dimensions sample_points sample_points_large

show_every – show output from every computation iteration.
If true, output is shown for each step. If false, output is only displayed when a minimum/maximum update occurs.
threads – number of threads to use.
If set to 0, it automatically uses the number of cores on the system.
dimensions – number of dimensions to test for.
This will typically be 2, 3, or 4.
sample_points – order of the number of points to sample.
This is essentially the number of sample points along a single edge of a simplex.
sample_points_large – order of the number of points to sample when a minimum/maximum update occurs.
When a minimum/maximum is updated, the same permutation of gradients is tested with a higher precision point density.