Earlier this week [1]NVIDIA confirmed some basic details around their next-gen Rosa CPU that succeeds Vera. Among the public confirmation was that it will feature a "Rigel" Armv9.2-A core iterating on their Olympus core design. With the basic details published, NVIDIA immediately [2]introduced Rigel core support into the GCC compiler . Now they have also upstreamed their initial Rigel core enablement into the LLVM Clang compiler.

Merged just in time for next week's LLVM 23 code branching is NVIDIA compiler engineers having landed their -mcpu=rigel targeting support for LLVM/Clang. This enablement for LLVM/Clang is similar to GCC in just adding the basic details and detection but not yet any specialized tuning / cost tables. It's good enough though to get "-mcpu=rigel" plumbed in and enabling all of the Rigel CPU core capabilities.

The NVIDIA Rigel CPU core support was added to LLVM/Clang Git via [3]this commit . With it having been merged in time for the LLVM/Clang 23.1 code branching next week, it will be found as part of the LLVM/Clang 23.1.0 stable release due out in late August.

It's great seeing NVIDIA getting this open-source upstreaming support underway for their Rosa CPU well in advance of its launch. Look out for more NVIDIA Rosa/Rigel CPU patches to come for the Linux kernel and other components over the coming months.



[1] https://www.phoronix.com/news/NVIDIA-Rosa-CPU-Rigel-Core

[2] https://www.phoronix.com/news/NVIDIA-Rigel-Upstream-GCC

[3] https://github.com/llvm/llvm-project/commit/061cd69c466ef49dfaafc81d79875cf2f7afa9fa