Operations with Arrays of Objects

Processing of matrix and vector arrays is a distinctive feature of IPP MX. Each Linear Algebra operation is implemented in the IPP MX interface with a base function, operating on single objects (matrices, vectors, or constants), and a collection of functions carrying out the same operation with one, two, or all source operands being arrays of matrices, vectors, or constants.

IPP MX base functions for matrices and vectors of certain small sizes (particularly, matrices of size 3x3, 4x4, 5x5, 6x6 and vectors of length 3, 4, 5, 6) are rather of supplementary importance, as their performance may sometimes be even worse than of a direct C code. The IPP MX interface mainly optimizes the use of functions implementing Linear Algebra operations with arrays of matrices, vectors and constants.

This section explains which operands composing source arrays of matrices, vectors, or constants such functions actually operate on and how the result is composed. Types of function operands are reflected in the function names (see the Function Naming section).

The discussion uses the following notation:

<ops>	An algebraic operation on matrices, vectors, and constants.
src, src1, src2	A single, first or second source vector, respectively.
dst	A destination vector.
src [i]	i-th element of the source vector.
dst [i]	i-th element of the destination vector.
src_j , src1_j , src2_j	j-th vector (matrix) in a single, first or second source array of vectors (matrices), respectively.
dst_j	j-th vector (matrix) in a destination array of vectors (matrices), respectively.
val	A constant.
Val_j	j-th constant in a source array of constants.
len	Vector length.
count	The number of matrices, vectors or constants in each array.