fir.hpp

Go to the documentation of this file.

#pragma once

#include <stdint.h>
#include <cublas_v2.h>


class FIR
{
    public:
    // M is decimation factor
    // f should be around 8
    // M * f is the length of the coefficient vector
    // nt is the number of input samples the fir will work on
    FIR(cublasHandle_t handle, cudaStream_t stream, float2 *hcoeff, int M, int f, int nt);
    ~FIR();

    cublasHandle_t _handle;
    cudaStream_t _stream;
    int _M;     // decimation factor
    int _f;     // taps per decimation factor
    int _ntap;  // M * f - number of taps in FIR filter
    int _nb;  // nt / M - output samples per input block
    int _nt;  // nb * M - input samples
    int _nout;  // f + nb - 1   // Length of output storage vector
    float2 *_dcoeff;  // f x M
    float2 *_dout; // nout
    float2 *_hout;  // nb
    float2 *_dtrapz; // nb x f

    void fir_apply(const float2 *din);
    void fir_shift();
    void fir_to_host(float2 *hout);
    void fir_to_dev(float2 *dout);

    void run_fir(const float2 *din, float2 *hout);

};