Work done so far on CUDA streams

[hercules2020/kcf.git] / src / fft_fftw.h

diff --git a/src/fft_fftw.h b/src/fft_fftw.h
index afe2c974205b7ebdf6e893b5a5fdb0b78dc5e613..2e60e02927825a83dffcc127d01c09129fed6056 100644 (file)
--- a/src/fft_fftw.h
+++ b/src/fft_fftw.h
@@ -20,12 +20,12 @@ class Fftw : public Fft
 {
 public:
     Fftw();
-    Fftw(int num_of_threads);
+    Fftw(unsigned num_of_threads);
     void init(unsigned width, unsigned height, unsigned num_of_feats, unsigned num_of_scales, bool big_batch_mode) override;
     void set_window(const cv::Mat & window) override;
-    void forward(const cv::Mat & real_input, ComplexMat & complex_result, float *real_input_arr) override;
-    void forward_window(std::vector<cv::Mat> patch_feats, ComplexMat & complex_result, cv::Mat & fw_all, float *real_input_arr) override;
-    void inverse(ComplexMat &  complex_input, cv::Mat & real_result, float *real_result_arr) override;
+    void forward(const cv::Mat & real_input, ComplexMat & complex_result, float *real_input_arr, cudaStream_t  stream) override;
+    void forward_window(std::vector<cv::Mat> patch_feats, ComplexMat & complex_result, cv::Mat & fw_all, float *real_input_arr, cudaStream_t stream) override;
+    void inverse(ComplexMat &  complex_input, cv::Mat & real_result, float *real_result_arr, cudaStream_t stream) override;
     ~Fftw() override;
 private:
     unsigned m_num_threads = 6;