Gladdy/EasyOpenCL

EasyOpenCL

No exposed low-level C, just the C++ standard library

Features

• Focus on programming the GPU instead of messing about with long C-style OpenCL function calls and manual memory management.
• CMake support for Linux and Mac - No more linking problems when you have installed the correct driver.
• Support for scalar values: pass additional structs to your kernel, eg. transformation matrices or custom constants.
• Chain kernels together in order to create a true pipeline on your GPU in which kernels can depend on multiple others. (example/main.cpp)
• Human readable OpenCL errors for easy debugging and teaching of the OpenCL basics.

Overview: it's this easy!

// example/simple.cpp
try {
  EasyOpenCL<float> framework (SHOW_DEBUG);
  
  //Register the kernel with the framework
  auto& square = framework.load("squarefloat");
  
  //Bind the input buffer (initial values from a std::vector) and the output buffer
  square.bindInput(0, std::vector<float> { 1.1, 2.2, 3.3, 4.4, 5.5, 11.0 });
  square.bindOutput(1);
  
  //Evaluate the kernel and display the output
  square.evaluate();
  square.showBuffer(1); // [ 1.21, 4.84, 10.89, 19.36, 30.25, 121 ]
} catch (std::exception& e) { std::cerr << "Error: " << e.what() << std::endl; }

// kernels/squarefloat.cl
__kernel void squarefloat(__global float* input, __global float* output)
{
  int i = get_global_id(0);
  output[i] = input[i] * input[i];
}

Getting started

All platforms:

• Update your graphics drivers
• Install the drivers with OpenCL support (NVIDIA CUDA Toolkit, AMD APP SDK or Intel OpenCL™ Code Builder)

git clone https://github.com/Gladdy/EasyOpenCL.git
cd EasyOpenCL
mkdir build && cd build
cmake ..
make
./simple

TODO:

• High priority

  • More examples - image processing, deep learning (matrix operations) and a renderer/raytracer
  • Asynchronous kernel calls + benchmarks of asynchronous vs synchronous kernel calls
  • Cleaning up the framework, getting public/private right + the different constructors

• Low priority:

  • Automatic generation of local and global work group sizes
  • Detect circular dependencies of kernels

• Mehh whenever I have the time:

  • .dot output of your current kernel dependency graph
  • Automatic generation of stub kernel (.cl) files based on the kernel links specified.

Thanks to:

• Dhruba Bandopadhyay: OpenCL Cookbook: Listing all devices and their critical attributes
• elhigu: cmake-findopencl
• Tom Scogland: OpenCL error checking