Asynchronous Linear Algebra Library

A cross-platform C++ library for running heavy linear algebra computations asynchronously on your GPU.
It provides a clean and intuitive interface using C++ operator overloading, while handling all OpenCL operations under the hood.

DNN::Matrix C(A + B);
C.waitForResults();
std::cout << C;

That’s it : asynchronous GPU matrix addition in just a few lines!

---

🚧 Project Status

The core async and OpenCL backend already works, but the library isn’t ready for real use yet, it currently just builds into a test executable. Here’s what’s coming next:

🧩 Short-Term Goals

• Refactor the code for easier extension and maintenance
• Add basic linear algebra operations (addition, multiplication, etc.)

🚀 Long-Term Goals

• Support more advanced operations (like eigenvalue computation)
• Optimize asynchronous performance
• Add support for other GPU APIs (e.g. CUDA)

---

⚙️ Build Instructions

This project depends on OpenCL, so make sure it’s installed first. A quick way to do this is via the Khronos OpenCL SDK.

The project builds with CMake and MSVC. In your CMakeLists.txt, point CMake to your OpenCL SDK with the OpenCLDir environment variable :

unset(OpenCL_DIR CACHE)
find_package(OpenCL PATHS "$ENV{OpenCLDir}" NO_DEFAULT_PATH CONFIG REQUIRED)

💡 Tip: For now, the easiest way to build and test the project is with CLion.