An implementation of two-dimensional separable and non-separable schemes for
discrete wavelet transform (DWT) with CDF 5/3 wavelet written in C
programming language using SIMD instructions.

The test suite contains four implementations:
	* non-separable discrete wavelet transform with barriers
	* non-separable discrete wavelet transform without barriers
	* separable discrete wavelet transform with barriers
	* separable discrete wavelet transform without barriers

Usage:
======

To get the results - optimal number of threads, optimal tile size and overall
performance - for your system just build the tests using make and then use
prepared scripts in tests folder.

Before running the scripts set OMP_PLACES and GOMP_CPU_AFFINITY in such a way
that new threads are started on separate physical cores first and after there
is no physical core without running thread, the hyperthreading is used.

To get the optimal number of threads run threads.sh in tests folder.

To get the optimal tile size run tile.sh in tests folder.

To get the overall performance of your system run image.sh in tests folder.

The results of all scripts are times in ns per pixel for each implementation
of DWT.

Acknowledgements
================

This software  has been supported by the Technology Agency of the Czech
Republic (TA CR) Competence Centres project V3C - Visual Computing Competence
Center (no. TE01020415).
