Intel® Integrated Performance Primitives (Intel® IPP) Developer Guide and Reference

ID 790148
Date 3/22/2024
Public
Document Table of Contents

Application Notes

  • The functions in this domain can be divided into two types: the functions that actually compress data, and transformation functions. The latters do not compress data but only modify them and prepare for further compression. The examples of such transformation are the Burrows-Weller Transform, or MoveToFront algorithm. To do data compression efficient, you should develop the proper consequence of functions of different type that will transform data and then compress them.
  • Compression ratio depends on the statistics of input data. For some types of input data no compression could be achieved at all.
  • The size of memory required for the output of data compression functions typically is not obvious. As a rule encoding functions uses less memory for output than the size of the input buffer, on the contrary decoding functions use more memory for output than the size of the input buffer. You should account these issues and allocate the proper quantity of output memory using the techniques provided by functions in this domain. For example, you can use a double-pointer technique for automatic shifting the user submitted pointer. For some other functions it is possible to compute the upper limit of the size of the required output buffer.