Fast double-precision vector and matrix maths library for Java, based around the concept of N-dimensional arrays.
This library is designed for use in games, simulations, raytracers, machine learning etc. where fast vector maths is important.
Some highlights:
- Vectorz can do over 1 billion 3D vector operations per second on a single thread.
- Specialised matrix types for efficient optimised operations (identity, diagonal, sparse etc.).
- Support for arbitrary
n-dimensional
numerical arrays.
Status
Vectorz is reasonably mature, battle tested and being used in production applications. The API is still evolving however as new features get added so you can expect a few minor changes, at least until version 1.0.0
Documentation
See the Vectorz Wiki
Example usage
Vector3 v=Vector3.of(1.0,2.0,3.0);
v.normalise(); // normalise v to a unit vector
Vector3 d=Vector3.of(10.0,0.0,0.0);
d.addMultiple(v, 5.0); // d = d + (v * 5)
Matrix33 m=Matrixx.createXAxisRotationMatrix(Math.PI);
Vector3 rotated=m.transform(d); // rotate 180 degrees around x axis
Key features
- Supports double typed vectors of arbitrary size
- Both mutable and immutable vectors are supported, enabling high performance algorithms
- Support for any size matrices, including higher dimensional (NDArray) matrices
- Ability to create lightweight view vectors (e.g. to access subranges of other vectors)
- Library of useful mathematical functions on vectors
- Vectors have lots of utility functionality implemented - Cloneable, Serializable, Comparable etc.
- Various specialised types of vectors/matrices types (e.g. identity matrices, diagonal matrices)
- Support for affine and other matrix transformations
- sparse arrays for space efficient large vectors and matrices where most elements are zero
- Operator system provides composable operators that can be applied to array elements
- Input / output of vectors and matrices - in various formats including readable edn format
Vectorz is deigned to allow the maximum performance possible for vector maths on the JVM.
This focus has driven a number of important design decisions:
- Support for sparse vectors and other specialised array types
- Specialised primitive-backed small vectors (1,2,3 and 4 dimensions) and matrices (1x1, 2x2, 3x3 and M*3)
- Abstract base classes preferred over interfaces to allow more efficient method dispatch
- Multiple types of vector are provided for optimised performance in special cases
- Hard-coded fast paths for most common 2D and 3D operations
- Vector operations are generally not thread safe, by design
- Concrete classes are generally final
If you have a use case that isn't yet well optimised then please post an issue - the aim is to make all common operations as efficient as possible.