cufft

I want to perform 441 2D, 32-by-32 FFTs using the batched method provided by the cuFFT library. The parameters of the transform are the following: int n[2] = {32,32}; int inembed[] = {32,32}; int onembed[] = {32,32/2+1}; cufftPlanMany(&plan,2,n,inembed,1,32*32,onembed,1,32*(32/2+1),CUFFT_D2Z,441); cufftPlanMany(&inverse_plan,2,n,onembed,1,32*32,inembed,1,32*32,CUFFT_Z2D,441); After I did the forward and inverse FFTs using the above plans, I could not get the original data back. Can anyone advise me how to set the parameters correctly for cudaPlanMany? Many thanks in advance. By the way, is it

I'm trying to find the fft of a dynamically allocated array. The input array is copied from host to device using cudaMemcpy2D . Then the fft is taken (cufftExecR2C) and the results are copied back from device to host. So my initial problem was how to use the pitch information in the fft. Then I found an answer here - CUFFT: How to calculate fft of pitched pointer? But unfortunately it doesn't work. The results I get are garbage values. Given below is my code. #define NRANK 2 #define BATCH 10 #include "cuda_runtime.h" #include "device_launch_parameters.h" #include <cufft.h> #include <stdio.h>

I am running CUFFT on chunks (N*N/p) divided in multiple GPUs, and I have a question regarding calculating the performance. First, a bit about how I am doing it: Send N*N/p chunks to each GPU Batched 1-D FFT for each row in p GPUs Get N*N/p chunks back to host - perform transpose on the entire dataset Ditto Step 1 Ditto Step 2 Gflops = ( 1e-9 * 5 * N * N *lg(N*N) ) / execution time and Execution time is calculated as: execution time = Sum(memcpyHtoD + kernel + memcpyDtoH times for row and col FFT for each GPU) Is this the correct way to evaluate CUFFT performance on multiple GPUs? Is there any

I'm trying to compute batch 1D FFTs using cufftPlanMany . The data set comes from a 3D field, stored in a 1D array, where I want to compute 1D FFTs in the x and y direction. The data is stored as shown in the figure below; continuous in x then y then z . Doing batch FFTs in the x -direction is (I believe) straighforward; with input stride=1 , distance=nx and batch=ny * nz , it computes the FFTs over elements {0,1,2,3} , {4,5,6,7} , ... , {28,29,30,31} . However, I can't think of a way to achieve the same for the FFTs in the y -direction. A batch for each xy plane is again straightforward