FFTAcceleration_scalBlocks_optFFT.h

/*
 * FFTAcceleration_scalBlocks_optFFT.h
 *
 *  Created on: Mar 16, 2016
 *      Author: gallardjm
 */
#ifndef FFTACC_SCALBLOCKS_OPTFFT_H_
#define FFTACC_SCALBLOCKS_OPTFFT_H_
 
#include "bhfmm/fft/FFTAccelerationImplementations/block/FFTDataContainer_scalBlocks.h"
#include "bhfmm/fft/FFTAccelerationAPI_extensions.h"
#include "bhfmm/fft/FFTSettings_preprocessor.h"
#include "bhfmm/fft/FFTSettings.h"
#include "bhfmm/fft/tools/optimizedFFT/optFFT_API_Factory.h"
 
/* 
 * Use FFTDataContainer_scalBlocks as Data container (2 arrays of matrices)
 */
class FFTAcceleration_scalBlocks_optFFT: public FFTAccelerationAPI_full {
 
public:
 
    FFTAcceleration_scalBlocks_optFFT(int order);
    ~~FFTAcceleration_scalBlocks_optFFT() {
        delete _optFFT_API;
        delete[] _blockSize;
        for (int i = 0; i < mardyn_get_max_threads(); ++i) {
            delete_matrix(_Re_tmp[i]);
            delete_matrix(_Im_tmp[i]);
        }
    }
 
    void FFT_initialize_Target(FFTAccelerableExpansion & Expansion);
    void FFT_initialize_Source(FFTAccelerableExpansion & Expansion,
            double radius);
    void FFT_initialize_TransferFunction(FFTAccelerableExpansion & Expansion);
 
    void FFT_finalize_Target(FFTAccelerableExpansion & Expansion,
            double radius);
 
    void FFT_M2L(FFTAccelerableExpansion & Source,
            FFTAccelerableExpansion & Target,
            FFTDataContainer* TransferFunction);
    void FFT_M2L_vec(FFTAccelerableExpansion & Source,
            FFTAccelerableExpansion & Target,
            FFTDataContainer* TransferFunction);
 
    //From FFTAcceleration_full
    void FFT_M2L_OrderReduction(FFTAccelerableExpansion & Source,
            FFTAccelerableExpansion & Target,
            FFTDataContainer* TransferFunction, int order);
    void FFT_M2L_OrderReduction_vec(FFTAccelerableExpansion & Source,
            FFTAccelerableExpansion & Target,
            FFTDataContainer* TransferFunction, int order);
 
    void FFT_M2L_2way(FFTAccelerableExpansion & Source1,
            FFTAccelerableExpansion & Source2,
            FFTAccelerableExpansion & Target1,
            FFTAccelerableExpansion & Target2,
            FFTDataContainer* TransferFunction);
    void FFT_M2L_2way_vec(FFTAccelerableExpansion & Source1,
            FFTAccelerableExpansion & Source2,
            FFTAccelerableExpansion & Target1,
            FFTAccelerableExpansion & Target2,
            FFTDataContainer* TransferFunction);
 
    void FFT_M2L_2way_ORed(FFTAccelerableExpansion & Source1,
            FFTAccelerableExpansion & Source2,
            FFTAccelerableExpansion & Target1,
            FFTAccelerableExpansion & Target2,
            FFTDataContainer* TransferFunction, int order);
    void FFT_M2L_2way_ORed_vec(FFTAccelerableExpansion & Source1,
            FFTAccelerableExpansion & Source2,
            FFTAccelerableExpansion & Target1,
            FFTAccelerableExpansion & Target2,
            FFTDataContainer* TransferFunction, int order);
 
protected:
    FFTDataContainer_scalBlocks* getFFTData(
            FFTAccelerableExpansion & Expansion);
    FFTDataContainer_scalBlocks* getFFTData_scal(
            FFTAccelerableExpansion & Expansion);
    optFFT_API* _optFFT_API;
 
    int _nbBlocks;
    int _nbLinePerBlock; //number of usefull line per block
    int* _blockSize;     //fft_ny size of each block
    //tmp matrices for FFT for each thread
    FFT_precision*** _Re_tmp;
    FFT_precision*** _Im_tmp;
 
    template<bool Vect, bool OrderRed>
    void FFT_M2L_template(FFTAccelerableExpansion & Source,
            FFTAccelerableExpansion & Target,
            FFTDataContainer* TransferFunction, int order);
    template<bool Vect, bool OrderRed>
    void FFT_M2L_2way_template(FFTAccelerableExpansion & Source1,
            FFTAccelerableExpansion & Source2,
            FFTAccelerableExpansion & Target1,
            FFTAccelerableExpansion & Target2,
            FFTDataContainer* TransferFunction, int order);
 
};
 
#endif