Merge pull request #28664 from pratham-mcw:armpl_dft_opt

Add ARMPL support for DFT Function #28664 - This PR introduces hal/armpl/ with implementation of 1D, 2D DFT and DCT routines using ARM Performance Libraries as a custom HAL replacement for OpenCV's DFT & DCT Function. - ArmPL MSI package is automatically downloaded and extracted via CMake when building on Windows ARM64, with a WITH_ARMPL option that defaults to ON for that platform. - Forward and inverse real DFT calls in dxt.cpp are routed through ArmPL when available, with scaling applied only when needed. - Test error thresholds in test_dxt.cpp are relaxed (from 1e-5 to 2e-4 for float ) to account for numerical differences between ArmPL and OpenCV's reference DFT results. **Performance Benchmarks :** <img width="993" height="835" alt="image" src="https://github.com/user-attachments/assets/76def647-6d20-4bce-8bc9-7363e723669f" /> - [x] I agree to contribute to the project under Apache 2 License. - [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV - [x] The PR is proposed to the proper branch
2026-07-29 23:33:05 +04:00 · 2026-05-14 13:26:21 +05:30
parent 357ad24262
commit eadaab5fbb
9 changed files with 3176 additions and 1 deletions
@@ -0,0 +1,91 @@
+#ifndef OPENCV_ARMPL_HAL_CORE_HPP
+#define OPENCV_ARMPL_HAL_CORE_HPP
+
+#ifdef HAVE_ARMPL
+
+#include <stddef.h>
+#include <fftw3.h>
+#include <opencv2/core/base.hpp>
+#include <opencv2/core/utility.hpp>
+
+#ifndef cvhalDFT
+struct cvhalDFT;
+#endif
+
+int armpl_hal_dftInit2D(cvhalDFT **context, int width, int height,
+                        int depth, int src_channels, int dst_channels,
+                        int flags, int nonzero_rows);
+
+int armpl_hal_dft2D(cvhalDFT *context, const unsigned char *src_data,
+                    size_t src_step, unsigned char *dst_data, size_t dst_step);
+
+int armpl_hal_dftFree2D(cvhalDFT *context);
+
+#undef  cv_hal_dftInit2D
+#define cv_hal_dftInit2D armpl_hal_dftInit2D
+
+#undef  cv_hal_dft2D
+#define cv_hal_dft2D armpl_hal_dft2D
+
+#undef  cv_hal_dftFree2D
+#define cv_hal_dftFree2D armpl_hal_dftFree2D
+
+struct ArmplDFTSpec_C_32fc {
+    fftwf_plan plan;
+    int        n;
+    bool       isInverse;
+};
+
+struct ArmplDFTSpec_C_64fc {
+    fftw_plan plan;
+    int       n;
+    bool      isInverse;
+};
+
+struct ArmplDFTSpec_R_32f {
+    fftwf_plan plan;
+    int        n;
+    bool       isInverse;
+    double     scale;
+};
+
+struct ArmplDFTSpec_R_64f {
+    fftw_plan plan;
+    int       n;
+    bool      isInverse;
+    double    scale;
+};
+
+int armpl_hal_dftInit1D(cvhalDFT **context, int len, int count,
+                        int depth, int flags, bool *needBuffer);
+
+int armpl_hal_dft1D(cvhalDFT *context,
+                    const unsigned char *src, unsigned char *dst);
+
+int armpl_hal_dftFree1D(cvhalDFT *context);
+
+#undef  cv_hal_dftInit1D
+#define cv_hal_dftInit1D armpl_hal_dftInit1D
+
+#undef  cv_hal_dft1D
+#define cv_hal_dft1D armpl_hal_dft1D
+
+#undef  cv_hal_dftFree1D
+#define cv_hal_dftFree1D armpl_hal_dftFree1D
+
+int armpl_hal_dctInit2D(cvhalDFT **context, int width, int height,
+                        int depth, int flags);
+int armpl_hal_dct2D(cvhalDFT *context, const unsigned char *src_data,
+                    size_t src_step, unsigned char *dst_data, size_t dst_step);
+int armpl_hal_dctFree2D(cvhalDFT *context);
+
+#undef  cv_hal_dctInit2D
+#define cv_hal_dctInit2D armpl_hal_dctInit2D
+#undef  cv_hal_dct2D
+#define cv_hal_dct2D armpl_hal_dct2D
+#undef  cv_hal_dctFree2D
+#define cv_hal_dctFree2D armpl_hal_dctFree2D
+
+#endif  // HAVE_ARMPL
+
+#endif  // OPENCV_ARMPL_HAL_CORE_HPP