linux下编译支持opencl的opencv for android

╰半夏微凉° 2021-12-21 10:29 286阅读 0赞

主要的步骤其他人已经写过，请参考这篇：[https://www.cnblogs.com/hrlnw/p/4720977.html][https_www.cnblogs.com_hrlnw_p_4720977.html]

操作的细节请参考附件的pdf:  https://files.cnblogs.com/files/ahfuzhang/opencvwithopencl4androidndk-141129030940-conversion-gate02.pdf.zip

用于测试的代码如下：

//jpg2gary.cpp
    #include <inttypes.h>
    #include <stdio.h>
    #include <stdlib.h>
    #include <string.h>
    #include <sys/time.h>
    #include <opencv2/core.hpp>
    #include <opencv2/opencv.hpp>
    #include <opencv2/core/ocl.hpp>
    #include <fstream>
    
    #ifndef P
    #define P(format, ...)                                                    \
        do {                                                                  \
            printf("%s %s %d " format "\n", __FILE__, __FUNCTION__, __LINE__, \
                   ##__VA_ARGS__);                                            \
            fflush(stdout);                                                   \
        } while (0);
    #endif
    
    void cpu(const char* img, int times) {
        cv::Mat image = cv::imread(img, cv::IMREAD_UNCHANGED);
        cv::Mat out;
        struct timeval start, end;
        gettimeofday(&start, NULL);
        for (int i = 0; i < times; i++) {
            cv::cvtColor(image, out, cv::COLOR_BGR2GRAY);
        }
        gettimeofday(&end, NULL);
        P("run times:%d, spend:%d ms", times, (end.tv_sec - start.tv_sec) * 1000 +
                                           (end.tv_usec - start.tv_usec) / 1000);
    }
    
    void opencl(const char* img, int times) {
        cv::Mat image = cv::imread(img, cv::IMREAD_UNCHANGED);
        //cv::UMat u_img = image.getUMat(cv::ACCESS_READ);
        cv::UMat u_img;
        image.copyTo(u_img);
        cv::UMat out;
        struct timeval start, end;
        gettimeofday(&start, NULL);
        for (int i = 0; i < times; i++) {
            cv::cvtColor(u_img, out, cv::COLOR_BGR2GRAY);
        }
        gettimeofday(&end, NULL);
        P("run times:%d, spend:%d ms", times, (end.tv_sec - start.tv_sec) * 1000 +
                                           (end.tv_usec - start.tv_usec) / 1000);
    }
    
    int init_col(){
        cv::ocl::setUseOpenCL(true);
      if (!cv::ocl::haveOpenCL()) {
        P("OpenCL is not available...");
        return -1;
      }
      cv::ocl::Context context;
      if (!context.create(cv::ocl::Device::TYPE_GPU)) {
        P("Failed creating the context...");
        return -1;
      }
      std::vector<cv::ocl::PlatformInfo> platform_info;
      cv::ocl::getPlatfomsInfo(platform_info);
      for (int i = 0; i < platform_info.size(); i++) {
        cv::ocl::PlatformInfo sdk = platform_info.at(i);
        for (int j = 0; j < sdk.deviceNumber(); j++) {
          cv::ocl::Device device;
          sdk.getDevice(device, j);
    
          std::cout << "\n\n*********************\n Device " << i + 1 << std::endl;
          std::cout << "Vendor ID: " << device.vendorID() << std::endl;
          std::cout << "Vendor name: " << device.vendorName() << std::endl;
          std::cout << "Name: " << device.name() << std::endl;
          std::cout << "Driver version: " << device.driverVersion() << std::endl;
          std::cout << "available: " << device.available() << std::endl;
    
          if (device.isAMD()) std::cout << "Is an AMD device" << std::endl;
          if (device.isIntel()) std::cout << "Is a Intel device" << std::endl;
    
          std::cout << "Global Memory size: " << device.globalMemSize()
                    << std::endl;
          std::cout << "Memory cache size: " << device.globalMemCacheSize()
                    << std::endl;
          std::cout << "Memory cache type: " << device.globalMemCacheType()
                    << std::endl;
          std::cout << "Local Memory size: " << device.localMemSize() << std::endl;
          std::cout << "Local Memory type: " << device.localMemType() << std::endl;
          std::cout << "Max Clock frequency: " << device.maxClockFrequency()
                    << std::endl;
        }
      }
      if (!cv::ocl::haveOpenCL()) {
        P("OpenCL is not available, again...");
        return -1;
      }  
      cv::ocl::Device(context.device(0));   
      return 0; 
    }
    
    int main(int argc, char* argv[]) {
        if (argc < 3) {
            printf("usage:%s <from> <cpu/opencl> [times=1]\n", argv[0]);
            return 0;
        }
        int times = 1;
        if (argc >= 4) {
            times = atoi(argv[3]);
        }
        if (strcmp(argv[2], "cpu") == 0) {
            cpu(argv[1], times);
        } else if (strcmp(argv[2], "opencl") == 0) {
            if (0!=init_col()){
                return 1;
            }
            opencl(argv[1], times);
        } else {
            P("unknow cpu/opencl");
            return 0;
        }
    
        return 1;
    }

使用xiaomi mix 2s, 高通骁龙 845， GPU Adreno 630, 对一张1080\*1443尺寸的图片使用cvtColor转换RGB到灰度。  
连续执行1000次：  
   CPU  595ms  
   OpenCL  96ms

加速6.2倍！

转载于:https://www.cnblogs.com/ahfuzhang/p/11071865.html

[https_www.cnblogs.com_hrlnw_p_4720977.html]: https://www.cnblogs.com/hrlnw/p/4720977.html