#include "fft_cufft.h"
-void cuFFT::init(unsigned width, unsigned height, unsigned num_of_feats, unsigned num_of_scales, bool big_batch_mode)
+cuFFT::cuFFT()
{
- m_width = width;
- m_height = height;
- m_num_of_feats = num_of_feats;
- m_num_of_scales = num_of_scales;
- m_big_batch_mode = big_batch_mode;
-
- std::cout << "FFT: cuFFT" << std::endl;
-
- if(m_height*(m_width/2+1) > 1024){
- std::cerr << "Image dimension after forward FFT are too big for CUDA kernels." << std::endl;
- std::exit(EXIT_FAILURE);
- }
-
- //FFT forward one scale
- {
- CudaSafeCall(cudaMalloc(&data_f, m_height*m_width*sizeof(cufftReal)));
-
- CufftErrorCheck(cufftPlan2d(&plan_f, m_height, m_width, CUFFT_R2C));
-
-
- }
- //FFT forward all scales
- if(m_num_of_scales > 1 && m_big_batch_mode)
- {
- CudaSafeCall(cudaMalloc(&data_f_all_scales, m_height*m_num_of_scales*m_width*sizeof(cufftReal)));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_scales;
- int idist = m_height*m_width, odist = m_height*(m_width/2+1);
- int istride = 1, ostride = 1;
- int *inembed = n, onembed[] = {(int)m_height, (int)m_width/2+1};
-
- CufftErrorCheck(cufftPlanMany(&plan_f_all_scales, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_R2C, howmany));
- }
- //FFT forward window one scale
- {
- CudaSafeCall(cudaHostAlloc(&data_fw, m_height*m_num_of_feats*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_fw_d, data_fw, 0));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_feats;
- int idist = m_height*m_width, odist = m_height*(m_width/2+1);
- int istride = 1, ostride = 1;
- int *inembed = n, onembed[] = {(int)m_height, (int)m_width/2+1};
+ CudaSafeCall(cudaSetDeviceFlags(cudaDeviceMapHost));
+ cudaErrorCheck(cublasCreate(&cublas));
+ cudaErrorCheck(cublasSetStream(cublas, cudaStreamPerThread));
+}
- CufftErrorCheck(cufftPlanMany(&plan_fw, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_R2C, howmany));
- }
- //FFT forward window all scales all feats
- if(m_num_of_scales > 1 && m_big_batch_mode)
- {
- CudaSafeCall(cudaHostAlloc(&data_fw_all_scales, m_height*m_num_of_feats*m_num_of_scales*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_fw_all_scales_d, data_fw_all_scales, 0));
+cufftHandle cuFFT::create_plan_fwd(uint howmany) const
+{
+ int rank = 2;
+ int n[] = {(int)m_height, (int)m_width};
+ int idist = m_height * m_width, odist = m_height * (m_width / 2 + 1);
+ int istride = 1, ostride = 1;
+ int *inembed = n, onembed[] = {(int)m_height, (int)m_width / 2 + 1};
+
+ cufftHandle plan;
+ cudaErrorCheck(cufftPlanMany(&plan, rank, n, inembed, istride, idist, onembed, ostride, odist, CUFFT_R2C, howmany));
+ cudaErrorCheck(cufftSetStream(plan, cudaStreamPerThread));
+ return plan;
+}
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_scales*m_num_of_feats;
- int idist = m_height*m_width, odist = m_height*(m_width/2+1);
- int istride = 1, ostride = 1;
- int *inembed = n, onembed[] = {(int)m_height, (int)m_width/2+1};
+cufftHandle cuFFT::create_plan_inv(uint howmany) const
+{
+ int rank = 2;
+ int n[] = {(int)m_height, (int)m_width};
+ int idist = m_height * (m_width / 2 + 1), odist = m_height * m_width;
+ int istride = 1, ostride = 1;
+ int inembed[] = {(int)m_height, (int)m_width / 2 + 1}, *onembed = n;
+
+ cufftHandle plan;
+ cudaErrorCheck(cufftPlanMany(&plan, rank, n, inembed, istride, idist, onembed, ostride, odist, CUFFT_C2R, howmany));
+ cudaErrorCheck(cufftSetStream(plan, cudaStreamPerThread));
+ return plan;
+}
- CufftErrorCheck(cufftPlanMany(&plan_fw_all_scales, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_R2C, howmany));
-
-
- }
- //FFT inverse one scale
- {
- CudaSafeCall(cudaHostAlloc(&data_i_features, m_height*m_num_of_feats*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_i_features_d, data_i_features, 0));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_feats;
- int idist = m_height*(m_width/2+1), odist = 1;
- int istride = 1, ostride = m_num_of_feats;
- int inembed[] = {(int)m_height, (int)m_width/2+1}, *onembed = n;
- CufftErrorCheck(cufftPlanMany(&plan_i_features, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_C2R, howmany));
- }
- //FFT inverse all scales
- if(m_num_of_scales > 1)
- {
- CudaSafeCall(cudaHostAlloc(&data_i_features_all_scales, m_height*m_num_of_feats*m_num_of_scales*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_i_features_all_scales_d, data_i_features_all_scales, 0));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_feats*m_num_of_scales;
- int idist = m_height*(m_width/2+1), odist = 1;
- int istride = 1, ostride = m_num_of_feats*m_num_of_scales;
- int inembed[] = {(int)m_height, (int)m_width/2+1}, *onembed = n;
+void cuFFT::init(unsigned width, unsigned height, unsigned num_of_feats, unsigned num_of_scales)
+{
+ Fft::init(width, height, num_of_feats, num_of_scales);
- CufftErrorCheck(cufftPlanMany(&plan_i_features_all_scales, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_C2R, howmany));
- }
- //FFT inverse one channel one scale
- {
- CudaSafeCall(cudaHostAlloc(&data_i_1ch, m_height*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_i_1ch_d, data_i_1ch, 0));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = 1;
- int idist = m_height*(m_width/2+1), odist = 1;
- int istride = 1, ostride = 1;
- int inembed[] = {(int)m_height, (int)m_width/2+1}, *onembed = n;
+ std::cout << "FFT: cuFFT" << std::endl;
- CufftErrorCheck(cufftPlanMany(&plan_i_1ch, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_C2R, howmany));
- }
- //FFT inverse one channel all scales
- if(m_num_of_scales > 1 && m_big_batch_mode)
- {
- CudaSafeCall(cudaHostAlloc(&data_i_1ch_all_scales, m_height*m_num_of_scales*m_width*sizeof(cufftReal), cudaHostAllocMapped));
- CudaSafeCall(cudaHostGetDevicePointer(&data_i_1ch_all_scales_d, data_i_1ch_all_scales, 0));
-
- int rank = 2;
- int n[] = {(int)m_height, (int)m_width};
- int howmany = m_num_of_scales;
- int idist = m_height*(m_width/2+1), odist = 1;
- int istride = 1, ostride = m_num_of_scales;
- int inembed[] = {(int)m_height, (int)m_width/2+1}, *onembed = n;
+ plan_f = create_plan_fwd(1);
+ plan_fw = create_plan_fwd(m_num_of_feats);
+ plan_i_1ch = create_plan_inv(1);
- CufftErrorCheck(cufftPlanMany(&plan_i_1ch_all_scales, rank, n,
- inembed, istride, idist,
- onembed, ostride, odist,
- CUFFT_C2R, howmany));
- }
+#ifdef BIG_BATCH
+ plan_f_all_scales = create_plan_fwd(m_num_of_scales);
+ plan_fw_all_scales = create_plan_fwd(m_num_of_scales * m_num_of_feats);
+ plan_i_all_scales = create_plan_inv(m_num_of_scales);
+#endif
}
-void cuFFT::set_window(const cv::Mat &window)
+void cuFFT::set_window(const MatDynMem &window)
{
- m_window = window;
+ Fft::set_window(window);
+ m_window = window;
}
-ComplexMat cuFFT::forward(const cv::Mat &input)
+void cuFFT::forward(const MatScales &real_input, ComplexMat &complex_result)
{
- ComplexMat complex_result;
- if(m_big_batch_mode && input.rows == (int)(m_height*m_num_of_scales)){
- CudaSafeCall(cudaMemcpy(data_f_all_scales, input.ptr<cufftReal>(), m_height*m_num_of_scales*m_width*sizeof(cufftReal), cudaMemcpyHostToDevice));
- complex_result.create(m_height, m_width / 2 + 1, m_num_of_scales);
- CufftErrorCheck(cufftExecR2C(plan_f_all_scales, reinterpret_cast<cufftReal*>(data_f_all_scales),
- complex_result.get_p_data()));
- } else {
- CudaSafeCall(cudaMemcpy(data_f, input.ptr<cufftReal>(), m_height*m_width*sizeof(cufftReal), cudaMemcpyHostToDevice));
- complex_result.create(m_height, m_width/ 2 + 1, 1);
- CufftErrorCheck(cufftExecR2C(plan_f, reinterpret_cast<cufftReal*>(data_f),
- complex_result.get_p_data()));
- }
-
- return complex_result;
+ Fft::forward(real_input, complex_result);
+ auto in = static_cast<cufftReal *>(const_cast<MatScales&>(real_input).deviceMem());
+
+ if (real_input.size[0] == 1)
+ cudaErrorCheck(cufftExecR2C(plan_f, in, complex_result.get_dev_data()));
+#ifdef BIG_BATCH
+ else
+ cudaErrorCheck(cufftExecR2C(plan_f_all_scales, in, complex_result.get_dev_data()));
+#endif
}
-ComplexMat cuFFT::forward_raw(float *input)
+void cuFFT::forward_window(MatScaleFeats &feat, ComplexMat &complex_result, MatScaleFeats &temp)
{
- ComplexMat dummy;
- return dummy;
-}
+ Fft::forward_window(feat, complex_result, temp);
-ComplexMat cuFFT::forward_window(const std::vector<cv::Mat> &input)
-{
- int n_channels = input.size();
- ComplexMat result;
- if(n_channels > (int) m_num_of_feats){
- cv::Mat in_all(m_height * n_channels, m_width, CV_32F, data_fw_all_scales);
- for (int i = 0; i < n_channels; ++i) {
- cv::Mat in_roi(in_all, cv::Rect(0, i*m_height, m_width, m_height));
- in_roi = input[i].mul(m_window);
- }
-
- result.create(m_height, m_width/2 + 1, n_channels,m_num_of_scales);
-
- CufftErrorCheck(cufftExecR2C(plan_fw_all_scales, reinterpret_cast<cufftReal*>(data_fw_all_scales_d), result.get_p_data()));
- } else {
- cv::Mat in_all(m_height * n_channels, m_width, CV_32F, data_fw);
- for (int i = 0; i < n_channels; ++i) {
- cv::Mat in_roi(in_all, cv::Rect(0, i*m_height, m_width, m_height));
- in_roi = input[i].mul(m_window);
+ cufftReal *temp_data = temp.deviceMem();
+ uint n_scales = feat.size[0];
+
+ for (uint s = 0; s < n_scales; ++s) {
+ for (uint ch = 0; ch < uint(feat.size[1]); ++ch) {
+ cv::Mat feat_plane = feat.plane(s, ch);
+ cv::Mat temp_plane = temp.plane(s, ch);
+ temp_plane = feat_plane.mul(m_window);
}
-
- result.create(m_height, m_width/2 + 1, n_channels);
-
- CufftErrorCheck(cufftExecR2C(plan_fw, reinterpret_cast<cufftReal*>(data_fw_d), result.get_p_data()));
}
- return result;
-}
-cv::Mat cuFFT::inverse(const ComplexMat &input)
-{
- int n_channels = input.n_channels;
- cufftComplex *in = reinterpret_cast<cufftComplex*>(input.get_p_data());
-
- if(n_channels == 1){
- cv::Mat real_result(m_height, m_width, CV_32FC1, data_i_1ch);
-
- CufftErrorCheck(cufftExecC2R(plan_i_1ch, in, reinterpret_cast<cufftReal*>(data_i_1ch_d)));
- cudaDeviceSynchronize();
-
- return real_result/(m_width*m_height);
- } else if(n_channels == (int) m_num_of_scales){
- cv::Mat real_result(m_height, m_width, CV_32FC(n_channels), data_i_1ch_all_scales);
-
- CufftErrorCheck(cufftExecC2R(plan_i_1ch_all_scales, in, reinterpret_cast<cufftReal*>(data_i_1ch_all_scales_d)));
- cudaDeviceSynchronize();
-
- return real_result/(m_width*m_height);
- } else if(n_channels == (int) m_num_of_feats * (int) m_num_of_scales){
- cv::Mat real_result(m_height, m_width, CV_32FC(n_channels), data_i_features_all_scales);
-
- CufftErrorCheck(cufftExecC2R(plan_i_features_all_scales, in, reinterpret_cast<cufftReal*>(data_i_features_all_scales_d)));
- cudaDeviceSynchronize();
-
- return real_result/(m_width*m_height);
- }
-
- cv::Mat real_result(m_height, m_width, CV_32FC(n_channels), data_i_features);
-
- CufftErrorCheck(cufftExecC2R(plan_i_features, in, reinterpret_cast<cufftReal*>(data_i_features_d)));
- cudaDeviceSynchronize();
-
- return real_result/(m_width*m_height);
+ if (n_scales == 1)
+ cudaErrorCheck(cufftExecR2C(plan_fw, temp_data, complex_result.get_dev_data()));
+#ifdef BIG_BATCH
+ else
+ cudaErrorCheck(cufftExecR2C(plan_fw_all_scales, temp_data, complex_result.get_dev_data()));
+#endif
}
-float* cuFFT::inverse_raw(const ComplexMat &input)
+void cuFFT::inverse(ComplexMat &complex_input, MatScales &real_result)
{
- cufftComplex *in = reinterpret_cast<cufftComplex*>(input.get_p_data());
-
- CufftErrorCheck(cufftExecC2R(plan_i_features_all_scales, in, reinterpret_cast<cufftReal*>(data_i_features_all_scales_d)));
-
- return data_i_features_all_scales;
+ Fft::inverse(complex_input, real_result);
+
+ uint n_channels = complex_input.n_channels;
+ cufftComplex *in = reinterpret_cast<cufftComplex *>(complex_input.get_dev_data());
+ cufftReal *out = real_result.deviceMem();
+ float alpha = 1.0 / (m_width * m_height);
+
+ if (n_channels == 1)
+ cudaErrorCheck(cufftExecC2R(plan_i_1ch, in, out));
+#ifdef BIG_BATCH
+ else
+ cudaErrorCheck(cufftExecC2R(plan_i_all_scales, in, out));
+#endif
+ cudaErrorCheck(cublasSscal(cublas, real_result.total(), &alpha, out, 1));
+ // The result is a cv::Mat, which will be accesses by CPU, so we
+ // must synchronize with the GPU here
+ CudaSafeCall(cudaStreamSynchronize(cudaStreamPerThread));
}
cuFFT::~cuFFT()
{
- CufftErrorCheck(cufftDestroy(plan_f));
- CufftErrorCheck(cufftDestroy(plan_f_all_scales));
- CufftErrorCheck(cufftDestroy(plan_fw));
- CufftErrorCheck(cufftDestroy(plan_fw_all_scales));
- CufftErrorCheck(cufftDestroy(plan_i_1ch));
- CufftErrorCheck(cufftDestroy(plan_i_1ch_all_scales));
- CufftErrorCheck(cufftDestroy(plan_i_features));
- CufftErrorCheck(cufftDestroy(plan_i_features_all_scales));
-
- CudaSafeCall(cudaFree(data_f));
- CudaSafeCall(cudaFree(data_f_all_scales));
- CudaSafeCall(cudaFreeHost(data_fw));
- CudaSafeCall(cudaFreeHost(data_fw_all_scales));
- CudaSafeCall(cudaFreeHost(data_i_1ch));
- CudaSafeCall(cudaFreeHost(data_i_1ch_all_scales));
- CudaSafeCall(cudaFreeHost(data_i_features));
- CudaSafeCall(cudaFreeHost(data_i_features_all_scales));
+ cudaErrorCheck(cublasDestroy(cublas));
+
+ cudaErrorCheck(cufftDestroy(plan_f));
+ cudaErrorCheck(cufftDestroy(plan_fw));
+ cudaErrorCheck(cufftDestroy(plan_i_1ch));
+
+#ifdef BIG_BATCH
+ cudaErrorCheck(cufftDestroy(plan_f_all_scales));
+ cudaErrorCheck(cufftDestroy(plan_fw_all_scales));
+ cudaErrorCheck(cufftDestroy(plan_i_all_scales));
+#endif
}