Simple CUDA file for benchmarking cuFFT plan with different signal size, inspired by https://github.com/drufat/cuda-examples

simple_cuda_fft.cu

// Simple benchmarking FFT with Cuda, building on https://github.com/drufat/cuda-examples
// compiles with nvcc -DSIGNAL_SIZE=2048 -m64 -gencode arch=compute_30,code=sm_30 -O2 -use_fast_math -I commoninclude -l cufft cudafft.cu -o cudaff
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>
#include <sys/time.h>

#include <cuda_runtime.h>
#include "helper_functions.h"
#include "helper_cuda.h"
#include <cufft.h>

void runTest(int argc, char** argv);
// Signal size
#ifndef SIGNAL_SIZE
  #define SIGNAL_SIZE 512
#endif

int main(int argc, char** argv)
{
  runTest(argc, argv);
}

//! Run and time a simple test for CUDA
void runTest(int argc, char** argv)
{
  StopWatchInterface *hTimer = NULL;
  sdkCreateTimer(&hTimer);
  struct timeval start, stop;
  double interval = 0.;
  int iterations = 1000;

  // Allocate host memory for the signal
  float2* h_signal = (float2*)malloc(sizeof(float2) * SIGNAL_SIZE);
  // Initalize the memory for the signal
  for (unsigned int i = 0; i < SIGNAL_SIZE; ++i) {
	 h_signal[i].x = rand() / (float)RAND_MAX;
	 h_signal[i].y = 0;
  }
  int mem_size = sizeof(float2) * SIGNAL_SIZE;

  // Allocate device memory for signal
  float2* d_signal;
  cudaMalloc((void**)&d_signal, mem_size);
  // Copy host memory to device
  cudaMemcpy(d_signal, h_signal, mem_size,
				 cudaMemcpyHostToDevice);

  // CUFFT plan
  cufftHandle plan;
  cufftPlan1d(&plan, mem_size, CUFFT_C2C, 1);

  for (int i = -1; i < iterations; i++) {
	 //i == -1 -- warmup iteration
	 if (i == 0) {
		checkCudaErrors(cudaDeviceSynchronize());
		sdkResetTimer(&hTimer);
		sdkStartTimer(&hTimer);
		gettimeofday(&start, NULL);
	 }
	 cufftExecC2C(plan, (cufftComplex *)d_signal, (cufftComplex *)d_signal, CUFFT_FORWARD);
  }
  checkCudaErrors(cudaDeviceSynchronize());
  sdkStopTimer(&hTimer);
  gettimeofday(&stop, NULL);
  double gpuTime = 0.001 * sdkGetTimerValue(&hTimer);
  printf("Time elapsed for %d  iterations : %.5f s (%g ms per iteration)\n", iterations, gpuTime, 1000.0*gpuTime/(double)iterations);
  interval =  (stop.tv_sec - start.tv_sec)*1000.0;
  interval += (stop.tv_usec - start.tv_usec)/1000.0;
  printf("Time elapsed for %d iterations: %g ms (%g ms per iterations)\n", iterations, interval, interval/(double)iterations);

  //Destroy CUFFT context
  cufftDestroy(plan);

  // cleanup memory
  free(h_signal);
  cudaFree(d_signal);
}