deepin-ocr/3rdparty/opencv-4.5.4/samples/hal/README.md
wangzhengyang 718c41634f feat: 切换后端至PaddleOCR-NCNN,切换工程为CMake
1.项目后端整体迁移至PaddleOCR-NCNN算法,已通过基本的兼容性测试
2.工程改为使用CMake组织,后续为了更好地兼容第三方库,不再提供QMake工程
3.重整权利声明文件,重整代码工程,确保最小化侵权风险

Log: 切换后端至PaddleOCR-NCNN,切换工程为CMake
Change-Id: I4d5d2c5d37505a4a24b389b1a4c5d12f17bfa38c
2022-05-10 10:22:11 +08:00

1.3 KiB

Custom HAL samples

Samples in this folder are intended to demonstrate functionality replacement mechanism in the OpenCV library.

The c_hal is the example of pure C replacement library with all functions returning error. It can be used to verify error handling in the function switching code.

The slow_hal contains naive C++ implementations of the element-wise logical array operations (and, or, xor, not) making them twice slower than the default.

Build custom HAL replacement library

  1. Create folder for build (for example <home-dir>/my-hal-build)
  2. Go to the created folder and run cmake: cmake <opencv-src>/samples/hal/slow_hal
  3. Run make

After build you will find static library in the build folder: libslow_hal.a

Build OpenCV with HAL replacement

  1. Create folder for build (for example <home-dir>/my-opencv-build)
  2. Go to the created folder and run cmake:
    cmake \
        -DOpenCV_HAL_DIR="<home-dir>/my-hal-build/" \
        <opencv-src>
    
  3. Run make (or make opencv_perf_core to build the demonstration test executable only)
  4. After build you can run the tests and verify that some functions works slower:
    ./bin/opencv_perf_core --gtest_filter=*bitwise_and*