Microsoft implements graphics server and GPU acceleration in WSL

Microsoft company announced on the implementation of significant improvements in the WSL subsystem (Windows Subsystem for Linux), which provides running Linux executable files on Windows:

  • Added by support for running Linux applications with a graphical interface, eliminating the need to use X servers from other companies. Support is implemented through GPU access virtualization.

    Microsoft implements graphics server and GPU acceleration in WSL

    An open driver has been prepared for the Linux kernel dxgkrnl, which exposes the /dev/dxg device with services replicating the WDDM D3DKMT of the Windows kernel. The driver establishes a connection to the physical GPU using the VM bus. Linux applications have the same level of GPU access as native Windows applications, without the use of resource sharing between Windows and Linux.

    Microsoft implements graphics server and GPU acceleration in WSL

    Moreover, for Linux, the libd3d12.so library is provided, which provides direct access to the Direct3D 12 graphics API and is built from the same code as the Windows d3d12.dll library. A simplified version of the dxgi API is also provided in the form of the DxCore library (libdxcore.so). The libd3d12.so and libdxcore.so libraries are proprietary and only come in binary builds (mounted in /usr/lib/wsl/lib) that are compatible with Ubuntu, Debian, Fedora, Centos, SUSE and other distributions based on Glibc.

    Microsoft implements graphics server and GPU acceleration in WSL

    OpenGL support in Mesa is provided through layer, which translates calls to the DirectX 12 API. The method for implementing the Vulkan API is still at the planning stage.

    Microsoft implements graphics server and GPU acceleration in WSL

  • Added support for computer computing on graphics cards, which allows you to use hardware acceleration for tasks such as machine learning and artificial intelligence. At the first stage, WSL environments will support CUDA and DirectML, running on top of the D3D12 API (for example, in a Linux environment, you can run TensorFlow with a DirectML backend). OpenCL support is possible through a layer that maps calls to the DX12 API.

    Microsoft implements graphics server and GPU acceleration in WSL

  • WSL installation will soon be supported with a simple "wsl.exe --install" command.

Source: opennet.ru

Add a comment