Matrix3D objects store an arbitrary sized 3D grid of arbitrary floating point data.

Several of these matrix operations are built on top of the BLAS library and may be hardware accelerated on certain processors if the platform provides a hardware accelerated BLAS library. In particular, AltiVec acceleration is provided on G4 an G5 processors with Mac OS X.

Hardware accelerated methods sometimes preform multiple operations simultaneously for the same computational cost. The add operation, for example, also scales one of the matrices being added. When using these methods, it is often beneficial to structure code to take advantage of all of the operations preformed. It is far more efficient to scale and add a matrix simultaneously using add than to first scale using scale and then add using add.

Technical note: matrix objects are implemented using single precision floats, while much of the rest of the breve simulation environment uses double precision floating point math.