Helper Functions for Selection of

Functions
static constexpr int	pcl::FFTConvolution::FasterThanNonseparableFilterSize (int numThreads)
static constexpr int	pcl::SeparableConvolution::FasterThanNonseparableFilterSize (int numThreads)
static constexpr int	pcl::FFTConvolution::FasterThanSeparableFilterSize (int numThreads)

Detailed Description

Convolution Algorithms

Function Documentation

FasterThanNonseparableFilterSize() [1/2]

static constexpr int pcl::FFTConvolution::FasterThanNonseparableFilterSize ( int  numThreads )

inlinestaticconstexpr

Returns the minimum filter size in pixels for which FFT-based two-dimensional convolution is consistently faster than nonseparable convolution on the current PixInsight/PCL platform, for the specified number of parallel execution threads.

The values returned by this function have been determined experimentally on reference hardware for optimized execution on machines and builds with and without AVX2/FMA3 processor instruction support.

Definition at line 342 of file FFTConvolution.h.

FasterThanNonseparableFilterSize() [2/2]

static constexpr int pcl::SeparableConvolution::FasterThanNonseparableFilterSize ( int  numThreads )

inlinestaticconstexpr

Returns the minimum filter size in pixels for which separable convolution is consistently faster than nonseparable convolution on the current PixInsight/PCL platform, for the specified number of parallel execution threads.

The values returned by this function have been determined experimentally on reference hardware for optimized execution on machines and builds with and without AVX2/FMA3 processor instruction support.

Definition at line 396 of file SeparableConvolution.h.

FasterThanSeparableFilterSize()

static constexpr int pcl::FFTConvolution::FasterThanSeparableFilterSize ( int  numThreads )

inlinestaticconstexpr

Returns the minimum filter size in pixels for which FFT-based two-dimensional convolution is consistently faster than separable convolution on the current PixInsight/PCL platform, for the specified number of parallel execution threads.

The values returned by this function have been determined experimentally on reference hardware for optimized execution on machines and builds with and without AVX2/FMA3 processor instruction support.

Definition at line 393 of file FFTConvolution.h.