To work more closely with Linux distributions, Nvidia has decided to release the GPU driver code for Linux as open source. This opening concerns kernel modules for Ampere and Turing GPUs.
Previously proprietary and closed source code for Nvidia’s GPU core modules has been released on Github. The first driver concerned is R515, under GPL/MIT license, accessible on this directory. “This release is an important step toward improving the experience of using Nvidia GPUs on Linux, for tighter integration with the operating system, and for developers to debug, integrate, and contribute back,” has explained the specialist in graphics and network accelerators.
As part of the announcement, the group said it is working with Canonical, Red Hat and SuSE to improve packaging, deployment and support models for their joint customers. Feedback from these vendors hailed the openness: “Nvidia’s open-source GPU core modules will simplify installations and increase security for Ubuntu users, whether they are AI/ML developers, gamers, or cloud users.” , commented in particular Cindy Goldberg, vice president of alliances at Canonical.
GPUs under Ampere and Turing architecture first
Open kernel modules support all Ampere and Turing server-bound GPUs. This is not yet the case for that of the GeForce and Workstation GPUs, the latter being in alpha. “We focused on testing on a wide variety of workloads to ensure feature and performance parity with the proprietary driver,” Nvidia said.
This concerns, for example, G-SYNC (simultaneous screen sharing) or ray tracing for the Vulkan API in RTX cards as well as the OptiX software acceleration engine. Or even other stacks such as the CUDA calculation execution platform or the set of OpenGL image calculation functions. But other features are also included, such as support for the DMA-BUF framework dedicated to buffer sharing between peripherals and subsystems, which takes on its full meaning with the Hopper architecture.
Patches to submit to Nvidia
It is possible to download the R515 development driver as part of CUDA Toolkit 11.7or from the page of download under the “beta” drivers. The data center R515 will follow in later versions, says Nvidia. Aware of possible maturity issues, the manufacturer offers users to upload them to its GitHub repository tracker or via its existing end-user support forum. Please report security issues via the channels listed in the GitHub repository security policy. “We encourage community submissions through pull requests on the GitHub page,” Nvidia announces. “Submitted fixes will be reviewed and, if approved, incorporated with any changes into a future driver release.”
We would like to say thanks to the writer of this post for this remarkable content
Nvidia releases open source GPU driver for Linux Computerworld
Find here our social media profiles and the other related pageshttps://go-dedicated.com/related-pages/