Question

每当我使用cuFFT绘制程序获得的值并将结果与Matlab的结果进行比较时，我得到相同形状的图形，并且最大值和最小值的值在相同点处得到。但是，cuFFT产生的值远大于Matlab产生的值。 Matlab代码是

fs = 1000;                              % sample freq
D = [0:1:4]';                           % pulse delay times
t = 0 : 1/fs : 4000/fs;                 % signal evaluation time
w = 0.5;                                % width of each pulse
yp = pulstran(t,D,'rectpuls',w);
filt = conj(fliplr(yp));
xx = fft(yp,1024).*fft(filt,1024);
xx = (abs(ifft(xx)));

和具有相同输入的CUDA代码如下：

cufftExecC2C(plan, (cufftComplex *)d_signal, (cufftComplex *)d_signal, CUFFT_FORWARD);
cufftExecC2C(plan, (cufftComplex *)d_filter_signal, (cufftComplex *)d_filter_signal,     CUFFT_FORWARD);
ComplexPointwiseMul<<<blocksPerGrid, threadsPerBlock>>>(d_signal, d_filter_signal, NX);
cufftExecC2C(plan, (cufftComplex *)d_signal, (cufftComplex *)d_signal, CUFFT_INVERSE);

cuFFT还执行1024点FFT，批量大小为2。

如果缩放系数为NX=1024，则值不正确。请告诉我该怎么做。

Answer 1

这是从未答复的清单中删除此问题的最新答案。

您没有提供足够的信息来诊断您的问题，因为您缺少指定设置cuFFT计划的方式。您甚至没有指定是否完全相同的形状用于Matlab和cuFFT的信号（因此您只需缩放）或者大约相同的形状。但是，让我提出以下两点意见：

yp向量包含4000个元素;与fft(yp,1024)的那个相反，您通过将信号截断为1024元素来执行FFT;
反向cuFFT不会按矢量元素的数量执行缩放。

为了方便起见（它可能对其他用户有用），我在下面报告一个简单的FFT-IFFT方案，该方案还包括使用CUDA Thrust库执行的缩放。

#include <cufft.h>
#include <thrust/host_vector.h>
#include <thrust/device_vector.h>

/*********************/
/* SCALE BY CONSTANT */
/*********************/
class Scale_by_constant
{
    private:
        float c_;

    public:
        Scale_by_constant(float c) { c_ = c; };

        __host__ __device__ float2 operator()(float2 &a) const
        {
            float2 output;

            output.x = a.x / c_;
            output.y = a.y / c_;

            return output;
        }

};

int main(void){

    const int N=4;

    // --- Setting up input device vector    
    thrust::device_vector<float2> d_vec(N,make_cuComplex(1.f,2.f));

    cufftHandle plan;
    cufftPlan1d(&plan, N, CUFFT_C2C, 1);

    // --- Perform in-place direct Fourier transform
    cufftExecC2C(plan, thrust::raw_pointer_cast(d_vec.data()),thrust::raw_pointer_cast(d_vec.data()), CUFFT_FORWARD);

    // --- Perform in-place inverse Fourier transform
    cufftExecC2C(plan, thrust::raw_pointer_cast(d_vec.data()),thrust::raw_pointer_cast(d_vec.data()), CUFFT_INVERSE);

    thrust::transform(d_vec.begin(), d_vec.end(), d_vec.begin(), Scale_by_constant((float)(N)));

    // --- Setting up output host vector    
    thrust::host_vector<float2> h_vec(d_vec);

    for (int i=0; i<N; i++) printf("Element #%i; Real part = %f; Imaginary part: %f\n",i,h_vec[i].x,h_vec[i].y);

    getchar();
}

Answer 2

通过引入cuFFT回调功能，cuFFT执行的逆FFT所需的归一化可以直接嵌入cufftExecC2C调用中，方法是将归一化操作定义为__device__函数。 / p>

除了cuFFT用户指南，有关cuFFT回调功能，请参阅

CUDA Pro Tip: Use cuFFT Callbacks for Custom Data Processing

以下是通过cuFFT回调实现IFFT规范化的示例。

    #include <stdio.h>
    #include <assert.h>

    #include "cuda_runtime.h"
    #include "device_launch_parameters.h"

    #include <cufft.h>
    #include <cufftXt.h>

    /********************/
    /* CUDA ERROR CHECK */
    /********************/
    #define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
    inline void gpuAssert(cudaError_t code, char *file, int line, bool abort=true)
    {
        if (code != cudaSuccess)
        {
            fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
            if (abort) exit(code);
        }
    }

    /*********************/
    /* CUFFT ERROR CHECK */
    /*********************/
    // See http://stackoverflow.com/questions/16267149/cufft-error-handling
    #ifdef _CUFFT_H_
    static const char *_cudaGetErrorEnum(cufftResult error)
    {
        switch (error)
        {
            case CUFFT_SUCCESS:
                return "CUFFT_SUCCESS";

            case CUFFT_INVALID_PLAN:
                return "CUFFT_INVALID_PLAN";

            case CUFFT_ALLOC_FAILED:
                return "CUFFT_ALLOC_FAILED";

            case CUFFT_INVALID_TYPE:
                 return "CUFFT_INVALID_TYPE";

            case CUFFT_INVALID_VALUE:
                return "CUFFT_INVALID_VALUE";

            case CUFFT_INTERNAL_ERROR:
                return "CUFFT_INTERNAL_ERROR";

            case CUFFT_EXEC_FAILED:
                return "CUFFT_EXEC_FAILED";

            case CUFFT_SETUP_FAILED:
                return "CUFFT_SETUP_FAILED";

            case CUFFT_INVALID_SIZE:
                return "CUFFT_INVALID_SIZE";

            case CUFFT_UNALIGNED_DATA:
                return "CUFFT_UNALIGNED_DATA";
        }

        return "<unknown>";
    }
    #endif

    #define cufftSafeCall(err)      __cufftSafeCall(err, __FILE__, __LINE__)
    inline void __cufftSafeCall(cufftResult err, const char *file, const int line)
    {
        if( CUFFT_SUCCESS != err) {
        fprintf(stderr, "CUFFT error in file '%s', line %d\n %s\nerror %d: %s\nterminating!\n",__FILE__, __LINE__,err, \
                                _cudaGetErrorEnum(err)); \
        cudaDeviceReset(); assert(0); \
        }
    }

    __device__ void IFFT_Scaling(void *dataOut, size_t offset, cufftComplex element, void *callerInfo, void *sharedPtr) {

        float *scaling_factor = (float*)callerInfo;

        float2 output;
        output.x = cuCrealf(element);
        output.y = cuCimagf(element);

        output.x = output.x / scaling_factor[0];
        output.y = output.y / scaling_factor[0];

        ((float2*)dataOut)[offset] = output;
}

    __device__ cufftCallbackStoreC d_storeCallbackPtr = IFFT_Scaling;

    /********/
    /* MAIN */
    /********/
    int main() {

        const int N = 16;

        cufftHandle plan;

        float2 *h_input             = (float2*)malloc(N*sizeof(float2));
        float2 *h_output1           = (float2*)malloc(N*sizeof(float2));
        float2 *h_output2           = (float2*)malloc(N*sizeof(float2));

        float2 *d_input;            gpuErrchk(cudaMalloc((void**)&d_input, N*sizeof(float2)));
        float2 *d_output1;          gpuErrchk(cudaMalloc((void**)&d_output1, N*sizeof(float2)));
        float2 *d_output2;          gpuErrchk(cudaMalloc((void**)&d_output2, N*sizeof(float2)));

        float *h_scaling_factor     = (float*)malloc(sizeof(float));
        h_scaling_factor[0] = 16.0f;
        float *d_scaling_factor;    gpuErrchk(cudaMalloc((void**)&d_scaling_factor, sizeof(float)));
        gpuErrchk(cudaMemcpy(d_scaling_factor, h_scaling_factor, sizeof(float), cudaMemcpyHostToDevice));

        for (int i=0; i<N; i++) {
            h_input[i].x = 1.0f;
            h_input[i].y = 0.f;
        }

        gpuErrchk(cudaMemcpy(d_input, h_input, N*sizeof(float2), cudaMemcpyHostToDevice));

        cufftSafeCall(cufftPlan1d(&plan, N, CUFFT_C2C, 1));

        cufftSafeCall(cufftExecC2C(plan, d_input, d_output1, CUFFT_FORWARD));
        gpuErrchk(cudaMemcpy(h_output1, d_output1, N*sizeof(float2), cudaMemcpyDeviceToHost));
        for (int i=0; i<N; i++) printf("Direct transform - %d - (%f, %f)\n", i, h_output1[i].x, h_output1[i].y);

        cufftCallbackStoreC h_storeCallbackPtr;
        gpuErrchk(cudaMemcpyFromSymbol(&h_storeCallbackPtr, d_storeCallbackPtr, sizeof(h_storeCallbackPtr)));

        cufftSafeCall(cufftXtSetCallback(plan, (void **)&h_storeCallbackPtr, CUFFT_CB_ST_COMPLEX, (void **)&d_scaling_factor));

        cufftSafeCall(cufftExecC2C(plan, d_output1, d_output2, CUFFT_INVERSE));
        gpuErrchk(cudaMemcpy(h_output2, d_output2, N*sizeof(float2), cudaMemcpyDeviceToHost));
        for (int i=0; i<N; i++) printf("Inverse transform - %d - (%f, %f)\n", i, h_output2[i].x, h_output2[i].y);

        cufftSafeCall(cufftDestroy(plan));

        gpuErrchk(cudaFree(d_input));
        gpuErrchk(cudaFree(d_output1));
        gpuErrchk(cudaFree(d_output2));

        return 0;
    }

修改

＆＃34;时刻＆＃34;执行回调操作由CUFFT_CB_ST_COMPLEX调用cufftXtSetCallback中指定。请注意，您可以使用相同的cuFFT计划加载和存储回调。

通过cuFFT在逆FFT中缩放

2 个答案: