windows OpenCL 环境配置

系统信息

键	值
GPU	Nvidia GeForce GTX 1650
CPU	AMD R5 4600H
operation system	版本 Windows 10 专业版 21H2

根据官网Conformant Products - The Khronos Group Inc
查询到硬件上对opencl的支持情况。查到GPU 1650支持opencl 3.0

弯路（只想配置环境可跳过）

KhronosGroup/OpenCL-SDK: OpenCL SDK (github.com)
使用vcpkg 安装opencl
先安装vcpkg vcpkg/README_zh_CN.md at 2022.04.12 · microsoft/vcpkg (github.com)

使用vcpkg 安装opencl

Package: opencl
Version: 2.2
Port-Version: 8
Architecture: x86-windows

安装提示信息如下

The package opencl is compatible with built-in CMake targets via CMake v3.6 and prior syntax

find_package(OpenCL REQUIRED)
target_link_libraries(main PRIVATE ${OpenCL_LIBRARIES})
target_include_directories(main PRIVATE ${OpenCL_INCLUDE_DIRS})

and the CMake v3.7 and beyond imported target syntax

find_package(OpenCL REQUIRED)
target_link_libraries(main PRIVATE OpenCL::OpenCL)

This package is only an OpenCL SDK. To actually run OpenCL code you also need to install an implementation.

WINDOWS: implementations typically ship with the drivers of you CPU/GPU vendors.
LINUX: implementations may be installed from your distro's repo or manually. There are too many to count.
APPLE: consult your distribution vendor on the state of OpenCL support: https://support.apple.com/en-us/HT202823

尝试了在CMake工程中使用vcpkg提供的库 - vcpkg_C++包管理器 - 博客园 (cnblogs.com)

这里出现了

PS D:\github_room\opencl\vcpkg> ./vcpkg integrate install
Applied user-wide integration for this vcpkg root.

All MSBuild C++ projects can now #include any installed libraries.
Linking will be handled automatically.
Installing new libraries will make them instantly available.

CMake projects should use: "-DCMAKE_TOOLCHAIN_FILE=D:/github_room/opencl/vcpkg/scripts/buildsystems/vcpkg.cmake"

在CMakeLists.txt中写

SET(CMAKE_TOOLCHAIN_FILE E:/vcpkg/clean/vcpkg/scripts/buildsystems/vcpkg.cmake)

这里提示，opencl的实现由cpu或gpu的厂商随驱动程序提供，那么需要找厂商文件或程序才是使用opencl的正途。

安装CUDA Toolkit

CUDA Toolkit | NVIDIA Developer
我选择了下载整个包之后安装。安装cuda 11.6版本。
安装好之后，已经提供了opencl环境。
如opencl头文件位置在

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.6\include\CL\cl.h

使用Clion

clion 版本为2022.1 配置mingw作为c++开发环境。
cmake 版本为3.23.1
一种偷懒而快速的方式，CodeBlocks会一路安装好mingw，把路径给clion也用，即可配置工具链。

代码文件与CMakeLists.txt

main.cpp参考OpenCL编译环境配置（VS+Nvidia） - 未雨愁眸 - 博客园 (cnblogs.com)
修改了两处

 if (strstr(ext_data, icd_ext) != NULL)

需要再引入一个头文件

#include <string.h>

第二个问题：c++出现中文乱码怎么解决？ - turtle的回答 - 知乎
https://www.zhihu.com/question/421326588/answer/1738490344
使用这个system("chcp 65001");

#include <stdio.h>  
#include <stdlib.h>  
#include <string.h>  
#include <iostream>  
#include <CL/cl.h>  
  
  
int main() {  
    system("chcp 65001");  
    //cl_platform 表示一个OpenCL的执行平台，关联到GPU硬件，如N卡，AMD卡  
    cl_platform_id *platforms;  
  
    //OpenCL中定义的跨平台的usigned int和int类型  
    cl_uint num_platforms;  
    cl_int i, err, platform_index = -1;  
  
    char *ext_data;  
    size_t ext_size;  
    const char icd_ext[] = "cl_khr_icd";  
  
    //要使platform工作，需要两个步骤。1 需要为cl_platform_id结构分配内存空间。2 需要调用clGetPlatformIDs初始化这些数据结构。一般还需要步骤0：询问主机上有多少platforms  
  
    //查询计算机上有多少个支持OpenCL的设备  
    err = clGetPlatformIDs(5, NULL, &num_platforms);  
    if (err < 0) {  
        perror("Couldn't find any platforms.");  
        exit(1);  
    }  
    printf("本机上支持OpenCL的环境数量: %d\n", num_platforms);  
  
    //为platforms分配空间  
    platforms = (cl_platform_id *)  
            malloc(sizeof(cl_platform_id) * num_platforms);  
  
    clGetPlatformIDs(num_platforms, platforms, NULL);  
  
    //获取GPU平台的详细信息  
    for (i = 0; i < num_platforms; i++) {  
        //获取缓存大小  
        err = clGetPlatformInfo(platforms[i],  
                                CL_PLATFORM_EXTENSIONS, 0, NULL, &ext_size);  
        if (err < 0) {  
            perror("Couldn't read extension data.");  
            exit(1);  
        }  
  
        printf("缓存大小: %zd\n", ext_size);  
  
        ext_data = (char *) malloc(ext_size);  
  
        //获取支持的扩展功能  
        clGetPlatformInfo(platforms[i], CL_PLATFORM_EXTENSIONS,  
                          ext_size, ext_data, NULL);  
        printf("平台 %d 支持的扩展功能: %s\n", i, ext_data);  
  
        //获取显卡的名称  
        char *name = (char *) malloc(ext_size);  
        clGetPlatformInfo(platforms[i], CL_PLATFORM_NAME,  
                          ext_size, name, NULL);  
        printf("平台 %d 是: %s\n", i, name);  
  
        //获取显卡的生产商名称  
        char *vendor = (char *) malloc(ext_size);  
        clGetPlatformInfo(platforms[i], CL_PLATFORM_VENDOR,  
                          ext_size, vendor, NULL);  
        printf("平台 %d 的生产商是: %s\n", i, vendor);  
  
        //获取平台版本  
        char *version = (char *) malloc(ext_size);  
        clGetPlatformInfo(platforms[i], CL_PLATFORM_VERSION,  
                          ext_size, version, NULL);  
        printf("平台 %d 的版本信息： %s\n", i, version);  
  
        //查询显卡是独立的还是嵌入的  
        char *profile = (char *) malloc(ext_size);  
        clGetPlatformInfo(platforms[i], CL_PLATFORM_PROFILE,  
                          ext_size, profile, NULL);  
        printf("平台 %d 是独立的（full profile）还是嵌入式的（embeded profile）?: %s\n", i, profile);  
  
        //查询是否支持ICD扩展  
        if (strstr(ext_data, icd_ext) != NULL)  
            platform_index = i;  
        std::cout << "平台ID = " << platform_index << std::endl;  
        /* Display whether ICD extension is supported */  
        if (platform_index > -1)  
            printf("平台 %d 支持ICD扩展: %s\n",  
                   platform_index, icd_ext);  
        std::cout << std::endl;  
  
        //释放空间  
        free(ext_data);  
        free(name);  
        free(vendor);  
        free(version);  
        free(profile);  
    }  
  
    if (platform_index <= -1)  
        printf("No platforms support the %s extension.\n", icd_ext);  
    getchar();  
  
    //释放资源  
    free(platforms);  
    return 0;  
}

CMakeLists.txt

cmake_minimum_required(VERSION 3.7)  
  
  
project (hello)   
  
add_executable(hello   
    main.cpp  
)  
  
find_package(OpenCL REQUIRED)  
target_link_libraries(hello PRIVATE OpenCL::OpenCL)

工具链为clion自动识别出的，

程序执行结果为

D:\github_room\testopencl\cmake-build-debug\hello.exe
Active code page: 65001
本机上支持OpenCL的环境数量: 2
缓存大小: 606
平台 0 支持的扩展功能: cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_ato
mics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_icd cl_
khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_d3d10_sharing cl_khr_d3d10_
sharing cl_nv_d3d11_sharing cl_nv_copy_opts cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics
cl_khr_device_uuid cl_khr_pci_bus_info cl_khr_external_semaphore cl_khr_external_memory cl_khr_external_semaphore_win32
cl_khr_external_memory_win32
平台 0 是: NVIDIA CUDA
平台 0 的生产商是: NVIDIA Corporation
平台 0 的版本信息： OpenCL 3.0 CUDA 11.6.134
平台 0 是独立的（full profile）还是嵌入式的（embeded profile）?: FULL_PROFILE
平台ID = 0
平台 0 支持ICD扩展: cl_khr_icd

缓存大小: 124
平台 1 支持的扩展功能: cl_khr_icd cl_khr_d3d10_sharing cl_khr_d3d11_sharing cl_khr_dx9_media_sharing cl_amd_event_callba
ck cl_amd_offline_devices
平台 1 是: AMD Accelerated Parallel Processing
平台 1 的生产商是: Advanced Micro Devices, Inc.
平台 1 的版本信息： OpenCL 2.1 AMD-APP (3075.13)
平台 1 是独立的（full profile）还是嵌入式的（embeded profile）?: FULL_PROFILE
平台ID = 1
平台 1 支持ICD扩展: cl_khr_icd


进程已结束,退出代码-1073741510 (0xC000013A: interrupted by Ctrl+C)