Software Technology

Data and Visualization

ZFP: Compressed Floating-Point Arrays

Principal Investigator: Peter Lindstrom, Lawrence Livermore National Laboratory

This project supports the development of zfp, which provides a compressed multidimensional array primitive for floating-point and integer data that supports very high-speed read and write random access. zfp’s tiled arrays allow the user to specify exactly how much storage to use on a per array or even per tile basis.  In addition to meeting storage constraints, zfp provides controls for satisfying absolute or relative error tolerances.  zfp is open source and is released under a BSD license.