我认为对您的问题的简短回答(使用单个cufftPlanMany
执行 3D 矩阵列的 1D FFT 的可能性)是否定的。
实际上,根据 执行的转换cufftPlanMany
,您称之为
cufftPlanMany(&handle, rank, n,
inembed, istride, idist,
onembed, ostride, odist, CUFFT_C2C, batch);
必须遵守高级数据布局。具体来说,一维 FFT 是根据以下布局计算出来的
input[b * idist + x * istride]
其中b
处理第b
-个信号,并且istride
是同一信号中两个连续项目之间的距离。如果 3D 矩阵具有维度M * N * Q
,并且如果您想沿列执行 1D 变换,则两个连续元素之间的距离将为M
,而两个连续信号之间的距离将为1
。此外,批处理执行的数量必须设置为等于M
。使用这些参数,您只能覆盖 3D 矩阵的一个切片。实际上,如果您尝试增加M
,那么 cuFFT 将开始尝试从第二行开始计算新的按列 FFT。这个问题的唯一解决方案是迭代调用cufftExecC2C
以覆盖所有Q
切片。
作为记录,以下代码提供了一个完整的示例,说明如何对 3D 矩阵的列执行 1D FFT。
#include <thrust/device_vector.h>
#include <cufft.h>
/********************/
/* CUDA ERROR CHECK */
/********************/
#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const 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);
}
}
int main() {
const int M = 3;
const int N = 4;
const int Q = 2;
thrust::host_vector<float2> h_matrix(M * N * Q);
for (int k=0; k<Q; k++)
for (int j=0; j<N; j++)
for (int i=0; i<M; i++) {
float2 temp;
temp.x = (float)(j + k * M);
//temp.x = 1.f;
temp.y = 0.f;
h_matrix[k*M*N+j*M+i] = temp;
printf("%i %i %i %f %f\n", i, j, k, temp.x, temp.y);
}
printf("\n");
thrust::device_vector<float2> d_matrix(h_matrix);
thrust::device_vector<float2> d_matrix_out(M * N * Q);
// --- Advanced data layout
// input[b * idist + x * istride]
// output[b * odist + x * ostride]
// b = signal number
// x = element of the b-th signal
cufftHandle handle;
int rank = 1; // --- 1D FFTs
int n[] = { N }; // --- Size of the Fourier transform
int istride = M, ostride = M; // --- Distance between two successive input/output elements
int idist = 1, odist = 1; // --- Distance between batches
int inembed[] = { 0 }; // --- Input size with pitch (ignored for 1D transforms)
int onembed[] = { 0 }; // --- Output size with pitch (ignored for 1D transforms)
int batch = M; // --- Number of batched executions
cufftPlanMany(&handle, rank, n,
inembed, istride, idist,
onembed, ostride, odist, CUFFT_C2C, batch);
for (int k=0; k<Q; k++)
cufftExecC2C(handle, (cufftComplex*)(thrust::raw_pointer_cast(d_matrix.data()) + k * M * N), (cufftComplex*)(thrust::raw_pointer_cast(d_matrix_out.data()) + k * M * N), CUFFT_FORWARD);
cufftDestroy(handle);
for (int k=0; k<Q; k++)
for (int j=0; j<N; j++)
for (int i=0; i<M; i++) {
float2 temp = d_matrix_out[k*M*N+j*M+i];
printf("%i %i %i %f %f\n", i, j, k, temp.x, temp.y);
}
}
当您想要对行执行一维变换时,情况就不同了。在这种情况下,两个连续元素之间1
的距离为 ,而两个连续信号之间的距离为M
。这允许您设置许多N * Q
转换,然后cufftExecC2C
只调用一次。作为记录,下面的代码提供了 3D 矩阵行的 1D 转换的完整示例。
#include <thrust/device_vector.h>
#include <cufft.h>
/********************/
/* CUDA ERROR CHECK */
/********************/
#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const 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);
}
}
int main() {
const int M = 3;
const int N = 4;
const int Q = 2;
thrust::host_vector<float2> h_matrix(M * N * Q);
for (int k=0; k<Q; k++)
for (int j=0; j<N; j++)
for (int i=0; i<M; i++) {
float2 temp;
temp.x = (float)(j + k * M);
//temp.x = 1.f;
temp.y = 0.f;
h_matrix[k*M*N+j*M+i] = temp;
printf("%i %i %i %f %f\n", i, j, k, temp.x, temp.y);
}
printf("\n");
thrust::device_vector<float2> d_matrix(h_matrix);
thrust::device_vector<float2> d_matrix_out(M * N * Q);
// --- Advanced data layout
// input[b * idist + x * istride]
// output[b * odist + x * ostride]
// b = signal number
// x = element of the b-th signal
cufftHandle handle;
int rank = 1; // --- 1D FFTs
int n[] = { M }; // --- Size of the Fourier transform
int istride = 1, ostride = 1; // --- Distance between two successive input/output elements
int idist = M, odist = M; // --- Distance between batches
int inembed[] = { 0 }; // --- Input size with pitch (ignored for 1D transforms)
int onembed[] = { 0 }; // --- Output size with pitch (ignored for 1D transforms)
int batch = N * Q; // --- Number of batched executions
cufftPlanMany(&handle, rank, n,
inembed, istride, idist,
onembed, ostride, odist, CUFFT_C2C, batch);
cufftExecC2C(handle, (cufftComplex*)(thrust::raw_pointer_cast(d_matrix.data())), (cufftComplex*)(thrust::raw_pointer_cast(d_matrix_out.data())), CUFFT_FORWARD);
cufftDestroy(handle);
for (int k=0; k<Q; k++)
for (int j=0; j<N; j++)
for (int i=0; i<M; i++) {
float2 temp = d_matrix_out[k*M*N+j*M+i];
printf("%i %i %i %f %f\n", i, j, k, temp.x, temp.y);
}
}