C语言版本的opencl的用法

本文介绍一下C语言使用Opencl的基本步骤，系统是windows10，IDE是Visual Studio2017。

step1：你需要安装一下opencl的windows版本，然后新建一个工程，在属性管理器里面给这个工程导入opencl的头文件。

 

step2：导入头文件

#include <iostream>
#include <fstream>
#include <sstream>
#include <CL/cl.h>
#include <opencv2/opencv.hpp>  //我这里的核函数操作的是图像，所以需要opencv的包
#include <opencv2/highgui.hpp>
#include <math.h>
#include <iostream>
#include <time.h>
#include <string>
using namespace cv;
using namespace std;
#pragma warning( disable : 4996 )

 

step3：获取平台和上下文

cl_context CreateContext()
{
	cl_int errNum;
	cl_uint numPlatforms;
	cl_platform_id firstPlatformId;
	cl_context context = NULL;

	//选择可用的平台中的第一个
	errNum = clGetPlatformIDs(1, &firstPlatformId, &numPlatforms);
	if (errNum != CL_SUCCESS || numPlatforms <= 0)
	{
		std::cerr << "Failed to find any OpenCL platforms." << std::endl;
		return NULL;
	}

	//创建一个OpenCL上下文环境
	cl_context_properties contextProperties[] =
	{
		CL_CONTEXT_PLATFORM,
		(cl_context_properties)firstPlatformId,
		0
	};
	context = clCreateContextFromType(contextProperties, CL_DEVICE_TYPE_GPU,
		NULL, NULL, &errNum);

	return context;
}

 

step4：创建设备并创建命令队列

cl_command_queue CreateCommandQueue(cl_context context, cl_device_id *device)
{
	cl_int errNum;
	cl_device_id *devices;
	cl_command_queue commandQueue = NULL;
	size_t deviceBufferSize = -1;

	// 获取设备缓冲区大小
	errNum = clGetContextInfo(context, CL_CONTEXT_DEVICES, 0, NULL, &deviceBufferSize);

	if (deviceBufferSize <= 0)
	{
		std::cerr << "No devices available.";
		return NULL;
	}

	// 为设备分配缓存空间
	devices = new cl_device_id[deviceBufferSize / sizeof(cl_device_id)];
	errNum = clGetContextInfo(context, CL_CONTEXT_DEVICES, deviceBufferSize, devices, NULL);

	//char name_data[100];
	//clGetDeviceInfo(devices[0], CL_DEVICE_NAME,sizeof(name_data), name_data, NULL);
	//printf("device:%s\n", name_data);
	//选取可用设备中的第一个
	commandQueue = clCreateCommandQueue(context, devices[0], 0, NULL);

	*device = devices[0];
	//delete[] devices;
	return commandQueue;
}

 

step5：构建并且编译program

cl_program CreateProgram(cl_context context, cl_device_id device, const char* fileName)
{
	cl_int errNum;
	cl_program program;

	std::ifstream kernelFile(fileName, std::ios::in);
	if (!kernelFile.is_open())
	{
		std::cerr << "Failed to open file for reading: " << fileName << std::endl;
		return NULL;
	}

	std::ostringstream oss;
	oss << kernelFile.rdbuf();

	std::string srcStdStr = oss.str();
	const char *srcStr = srcStdStr.c_str();
	program = clCreateProgramWithSource(context, 1,
		(const char**)&srcStr,
		NULL, NULL);

	errNum = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);

	return program;
}

 

step6：创建kernel

// 一、选择OpenCL平台并创建一个上下文
cl_context context = CreateContext();

// 二、 创建设备并创建命令队列
cl_command_queue commandQueue = CreateCommandQueue(context, &device);
cl_event histEvent = 0;

//创建和构建程序对象
cl_program program = CreateProgram(context, device, "main.cl");
cl_kernel kernel = clCreateKernel(program, "kernel_rgb2gray", NULL);

 

step7：新建文件main.cl，在里面新建一个核函数

__kernel void kernel_rgb2gray(__global unsigned char * rgbImage, 
                              __global int* result)
{
    int x = get_global_id(0);
    int y = get_global_id(1);
    int index = x * height + y;
    result[index] = rgbImage[index];
}

 

step8：为核函数设置参数

Mat image = imread("D://b.jpg");
int imgSize = image.rows * image.cols;
Mat srcImage;
cvtColor(image, srcImage, CV_BGR2GRAY);
cl_mem memResult = clCreateBuffer(context, CL_MEM_WRITE_ONLY,
		sizeof(int)*imgSize, NULL, NULL);
cl_mem memRgbImage = clCreateBuffer(context, CL_MEM_USE_HOST_PTR,
		sizeof(uchar) * 3 * imgSize, srcImage.data, NULL); 
errNum = clSetKernelArg(kernel, 0, sizeof(cl_mem), &memRgbImage);
errNum = clSetKernelArg(kernel, 1, sizeof(cl_mem), &memResult);

 

step9：执行核函数

int img_h = img.rows;
int img_w = img.cols;
size_t globalThreads[2] = { img_w, img_h };
//执行内核函数
errNum = clEnqueueNDRangeKernel(commandQueue, kernel, 2, 0, globalThreads, NULL,
		0, NULL, &histEvent);

 

step10：获取执行结果

int result* = new int[imgSize];
errNum = clEnqueueReadBuffer(commandQueue, memResult, CL_TRUE, 0,
		sizeof(float)*img_h*img_w, result, 1, &histEvent, NULL);