On May 11, NVIDIA announced that it has officially open-sourced its Linux GPU kernel module with dual licenses of GPL and MIT. Developers can find the relevant source code of the kernel module in the NVIDIA Open GPU Kernel Modules repo on GitHub.

GitHub address: https://github.com/NVIDIA/open-gpu-kernel-modules

Nvidia said the open source will improve the experience of using Nvidia GPUs in Linux systems, make the hardware and system more closely linked, and allow developers to debug, integrate and contribute. For Linux distributors, open source kernel modules add ease of use and improve the out-of-the-box user experience for signing and distributing NVIDIA GPU drivers. Canonical and SUSE can immediately package open source kernel modules with Ubuntu and SUSE Linux Enterprise Distributions.

However, this open source Linux GPU kernel module also has shortcomings. Nvidia said: "The current code base does not meet the Linux kernel design requirements, nor is it a candidate for Linux upstream. In the future, it will plan to cooperate with the Linux kernel community and partners (such as Canonical, Red Hat and SUSE) working together to develop a methodology for upstreaming.”

Supported Features

The first version of NVIDIA's open-source GPU kernel module is R515, which provides a complete build and packaged version of the driver in addition to the release of the source code.

  • The code is production ready for data center GPUs in the NVIDIA Turing and NVIDIA Ampere architecture families.
  • Support for GeForce and Workstation GPUs is high, GeForce and Workstation users can use this driver on NVIDIA Turing, NVIDIA Ampere architecture GPUs to run Linux desktops, and Vulkan and NVIDIA OptiX for multi-monitor, G-SYNC, and NVIDIA RTX ray tracing and more.
  • Users with Turing and Ampere GPUs can choose which modules to install. Users of pre-Turing hardware will continue to run closed source modules.

Note: All components of the open source kernel module driver stack must match the version in the distribution. For example, users cannot release source code, build or run it using previous or future versions of the user-mode stack.

Although there are shortcomings that cannot be upstreamed, this news is enough to cause a sensation in the developer community, surprising and delighting many industry insiders and developers.

Mike McGrath, VP of Linux Engineering at Red Hat: "Red Hat has worked with NVIDIA for many years, and we're excited to see them take the next step. We look forward to bringing these capabilities to our customers and improving interoperability with NVIDIA hardware.

Cindy Goldberg, Vice President, Canonical Chip Alliance: "As developers of Ubuntu, the most popular Linux-like operating system for developers, we are now able to tightly integrate Ubuntu with NVIDIA GPUs to provide developers with better access to cutting-edge fields such as AI and ML. support."

Markus Noga, General Manager, Business Critical Linux, SUSESUSE: "We at SUSE are delighted to see NVIDIA's decision to release the GPU kernel-mode driver as open source. This is a true milestone for the open source community and accelerated computing."

At present, NVIDIA's open-source Linux GPU kernel module has received 9k+ Stars on GitHub, and some developers expressed their support for NVIDIA's open-source move, and even bluntly said, "This is the most recent hardware support for open-source operating systems in more than a decade. One of the important things."

What are your thoughts on Nvidia's move to source Linux GPU kernel modules?

Welcome to leave a message and interact in the comment area~

For more open source details, see:

https://developer.nvidia.com/blog/nvidia-releases-open-source-gpu-kernel-modules/


六一
556 声望347 粉丝

SegmentFault 新媒体运营