Data Representations and Function Variants
Because different game engines may use very different representations of character animation data, HumanIK offers the choice of using several different data representations for 3D coordinates, orientations and scaling values.
The HumanIK functions that set and retrieve data fall into two main categories:
- Functions that represent translation, orientation (typically expressed as a quaternion) and scaling values in separate arrays of numbers. See also T, Q and S Values.
- Functions that represent translation, orientation and scaling values using a single 4x4 transform matrix. See also Transform Matrices.
These approaches are essentially equivalent; you can use whichever fits more easily into your project and your existing data structures. However, because HumanIK stores values internally using separate translation, quaternion rotation and scaling values, the functions that accept and return Node and Effector data as transformation matrices need to convert the values internally. These matrix functions are therefore less efficient than the functions that represent translation, orientation and scaling values in separate arrays. Although this difference in performance is extremely small, it may be non-negligeable when setting or retrieving values for many different Nodes or Effectors at the same time.
Function variants
Many functions in HumanIK, especially those that set and retrieve coordinate data for Effectors and Nodes, are offered in several variants that accept different types of arguments. You can choose the variant that best matches the requirements of your game engine, and that results in the best performance in your environment.
Many functions are provided in different versions that accept either floating-point numbers or double-precision numbers, as indicated by the following suffixes:
- fv: parameters must be floating-point numbers.
- dv: parameters must be double-precision numbers. These functions are intended for use
where floating-point values are unavailable. However, as these functions convert the
numeric values internally before processing, they are slower than their fv counterparts.
The following variants are listed in descending order of performance, starting with the fastest functions. The suffix of each function indicates its requirements.
- TQAlign: Sets and retrieves only translation and quaternion rotation, where both types of data are stored at specified offsets within a single array of numbers, aligned on 16 bytes. For example: {tx,ty,tz,tw,qx,qy,qz,qw}.
- TQSAlign: Sets and retrieves translation, quaternion rotation and uniform scale, where all types of data are stored at specified offsets within a single array of numbers, aligned on 16 bytes. For example: {tx,ty,tz,tw,qx,qy,qz,qw,sx,sy,sz,sw}.
- Q: Sets and retrieves only quaternion rotation.
- TQ: Sets and retrieves only translation and quaternion rotation.
- TQS: Sets and retrieves translation, quaternion rotation and uniform scale independently, where each type of data is stored in a separate array of numbers.
- TQNonUniformS: Sets and retrieves translation, quaternion rotation and non-uniform scale independently, where each type of data is stored in a separate array of numbers. For use in situations where the X, Y and Z axes use different scaling values.
For details on individual functions, see their descriptions in the API Reference.
Topics in this section
Except where otherwise noted, this work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License. Please see the Autodesk Creative Commons FAQ for more information.