dsp::FilterbankEngineCPU Class Reference

Discrete convolution filterbank step for CPU. More...

#include <FilterbankEngineCPU.h>

Inherits dsp::Filterbank::Engine.

Public Member Functions

 FilterbankEngineCPU ()
 Default Constructor.
 
void setup (dsp::Filterbank *)
 
void set_scratch (float *)
 
void perform (const dsp::TimeSeries *in, dsp::TimeSeries *out, uint64_t npart, uint64_t in_step, uint64_t out_step)
 
void perform (const dsp::TimeSeries *in, dsp::TimeSeries *out, dsp::TimeSeries *zero_DM_out, uint64_t npart, uint64_t in_step, uint64_t out_step)
 
void finish ()
 
FTransform::Planget_forward ()
 
FTransform::Planget_backward ()
 

Protected Attributes

FTransform::Planforward
 
FTransform::Planbackward
 
bool real_to_complex
 Complex-valued data.
 
float * scratch
 device scratch sapce
 
float * freq_domain_scratch [3]
 scratch space for forward fft This is an array of float pointers because we might be dealing not only with multiple polarizations, but also with cross polarization.
 
float * time_domain_scratch
 scratch space for backward fft
 
float * windowed_time_domain_scratch
 scratch space for apodization operation
 
const dsp::Responseresponse
 response kernel, from Filterbank
 
const dsp::Responsezero_DM_response
 zero DM response, from Filterbank
 
const dsp::Apodizationtemporal_apodization
 apodization kernel, from Filterbank
 
const dsp::Apodizationspectral_apodization
 
dsp::Responsepassband
 passband kernel, from Filterbank
 
bool matrix_convolution
 Whether or not to do matrix convolution, from Filterbank.
 
unsigned nchan_subband
 number of output channels per input channel
 
unsigned freq_res
 frequency resolution of response (dsp::Response::get_ndat)
 
unsigned nfilt_pos
 positive impulse from response (dsp::Response::get_impulse_pos)
 
unsigned nfilt_neg
 negative impulse from response (dsp::Response::get_impulse_neg)
 
uint64_t nsamp_fft
 number of samples in forward fft
 
unsigned nkeep
 number of samples to keep from each input sample. More...
 
bool verbose
 

Detailed Description

Discrete convolution filterbank step for CPU.

Member Data Documentation

◆ nkeep

unsigned dsp::FilterbankEngineCPU::nkeep
protected

number of samples to keep from each input sample.

This is essentially the number of fft points minus the total smearing from the response


The documentation for this class was generated from the following files:

Generated using doxygen 1.8.17