Function Naming: Small Matrix DomainTake look at the name of a random Intel® Integrated Performance Primitive (Intel® IPP) function for small matrix processing and see how much it tells you. Click on the components that make up the name of the function to learn what they signify.
The function prefix first tells you that this particular function belongs to the Intel IPP package and, more specifically, to the small matrix domain, where expected data is twodimensional matrix. The function base name and modifier suggest that in this case the function performs subtraction.
Object types describe
the type of objects passed to a function for processing and may be the
following:
The objects are given in the <object1><object2> format. In the given example vac means that a constant is subtracted from an array of vectors.
Currently the Intel IPP functions for small matrix processing perform operations on 32f (32bit floating point) and 64f (64bit double precision) data. This part of the function
name specifies the size of vectors or matrices on which the function
operates. The following sizes are used to specify the size of vectors:
The size of matrices can be indicated as
If this component is missing from a function name, this function is written for Nx1 (vector) or NxM (matrix) cases.
Finally, the descriptor
field further describes the data associated with the operation. It may
contain implied parameters and/or indicate additional required parameters.
To minimize the number of code branches in the function and thus reduce
potentially unnecessary execution overhead, most of the general functions
were split into separate primitive functions, with some of their parameters
entering the primitive function name as descriptors.
As you see, in this example the objects are defined by the Lpointer. The following descriptors are used in signal processing functions:
