Easier to install, Windows 10’s Linux subsystem makes better use of the GPU
In a first version of its WSL 2 subsystem, Microsoft wanted to simplify its installation and make it possible to control more workloads by making the most of the graphics circuits.
Microsoft has updated its Windows Subsystem for Linux 2 (WSL) to a preview version of Windows 10, whose features are accessible to novices and experienced developers alike. WSL2 can now perform compute functions on the GPU accelerator, including using Nvidia’s CUDA technology. This development was one of the promises made by Microsoft at the Build 2020 conference last May. WSL improvements are part of Windows 10 Build 20150 for Insider users. Formerly known as the Fast Ring, the Dev Channel is dedicated to testing features that aren’t necessarily tied to an upcoming version of Windows 10.
As the name suggests, Windows Subsystem for Linux 2 allows you to run a Linux kernel from Windows. WSL2 isn’t as good as a full-fledged Linux distribution running natively on the hardware, but it comes close. WSL2 is also not natively integrated into Windows. This is an option that can be enabled using the “Turn Windows features on and off” control panel in Windows 10, which will then download the necessary files and require restarting the PC.
One of the latest features in Build 20150 tries to make things even simpler. By typing wsl.exe -install as a command, Windows will take care of all the configuration. It is, however, not entirely clear if this command can be executed from the Command Line application, or if it requires the PowerShell interface. According to a blog post by Craig Loewen, Windows Program Manager, Microsoft plans to support automatic installation of a Linux distribution in the WSL soon.
Similarly, users can type wsl.exe -update to update the Linux kernel; wsl.exe -update -status to check kernel status; and wsl.exe -update -rollback to revert to a previous kernel version. According to Loewen, Linux kernel updates will actually be handled by Windows Update and automatically downloaded to your PC. The commands still exist to ensure precise control of the process.
GPU Computing Support
For the more adventurous looking to delve deeper into Linux, Microsoft has now implemented GPU computing support, which has traditionally been supported by hardware (and Linux) for nearly a decade. This announcement is just one of many more to come soon. “This release will also support artificial intelligence (AI) and machine learning (ML) workflows, allowing professionals and students alike to manage ML training workloads across all GPUs in the Windows ecosystem,” the company said in a blog post.
Specifically, Microsoft has partnered with Nvidia for a beta version of CUDA for WSL2, which includes machine learning support for popular ML tools, libraries, and frameworks, including PyTorch and TensorFlow. Microsoft also provides a preview of TensorFlow with a DirectML backend. AMD also showed a preview of the graphics driver unlocking DirectX 12 compatible GPU acceleration within the WSL on a wide range of its Radeon hardware.