Easy Computing Library v2

Overview

EasyCL v2 is open-source (GPL v3) header-only wrapper above OpenCL library. It's designed for easy and convenient use, allowing you to quickly write the host part of the program and flexibly use OpenCL kernels.

The library is built on the original OpenCL 2.0 C Specification. It allows you to bypass some of the inconveniences of the original OpenCL, providing work with abstractions of different levels, provides you to bypass the restriction of hard binding.

First version of this library was made for C++.

Installation

1. Install OpenCL library on your system.
2. Clone the repo $ git clone https://github.com/architector1324/EasyCL2.
3. Copy easycl.h to your project

Hello, World

1. Copy easycl.h to project folder
2. Create main.c:

#include <stdio.h>
#include "easycl.h"


int main() {
    // setup program and kernel
    EclProgram_t prog = {};
    eclProgramLoad("main.cl", &prog);

    EclKernel_t kern = {.name = "test"};

    // get platform, setup the computer
    EclPlatform_t plat = {};
    eclGetPlatform(0, &plat);

    EclComputer_t gpu = {};
    eclComputer(0, ECL_DEVICE_GPU, &plat, &gpu);

    // setup data container
    ecl_array(int, a, 12, {}, ECL_BUFFER_WRITE); // ecl::array<int> a[12] = {};
    int b = 5;

    // setup compute frame
    EclFrame_t frame = {
        .prog = &prog,
        .kern = &kern,
        .args = {{ECL_ARG_BUFFER, &a}, {ECL_ARG_VAR, &b, sizeof(int)}},
        .argsCount = 2
    };

    // 12 compute units combined into 3 groups
    eclComputerSend(&a, &gpu, ECL_EXEC_SYNC);
    eclComputerGrid(&frame, (EclWorkSize_t){.dim = 1, .sizes={12}}, (EclWorkSize_t){.dim = 1, .sizes={3}}, &gpu, ECL_EXEC_SYNC);
    eclComputerReceive(&a, &gpu, ECL_EXEC_SYNC);

    // output
    for(size_t i = 0; i < 12; i++)
        printf("%d\n", ((const int*)a.data)[i]);

    // clean resources
    eclBufferClear(&a);

    eclComputerClear(&gpu);
    eclPlatformClear(&plat);

    eclKernelClear(&kern);
    eclProgramClear(&prog);

    return 0;
}

3. Create main.cl:

kernel void test(global int* a, int b){
    size_t i = get_global_id(0);
    a[i] = b * ((int)get_group_id(0) + 1);
}

4. Type in terminal:

$ gcc -O3 -lOpenCL main.c -o a.out
$ ./a.out

Output:

5
5
5
10
10
10
15
15
15
20
20
20

If you have any questions, feel free to contact me olegsajaxov@yandex.ru