LLVM/Clang 23.1: AMD Zen 6 support, NVIDIA Rigel cores and partial C++26

The LLVM project has shipped version 23.1 on schedule — the first stable release of the LLVM 23 series and the planned update to the open compiler stack for the second half of 2026. As is customary for the project, the release brings a long list of changes affecting support for new architectures, language standards and tooling.

New processors

The headline addition is support for the AMD Zen 6 microarchitecture, code-named “znver6.” Along with it, the compiler gained AVX-512 BMM instructions. There is also initial support for the NVIDIA Rigel core, which will form the basis of the NVIDIA Rosa processor, as well as a scheduling model for the Arm C1-Ultra.

The list of target architectures grew in other directions too: support was added for AMDGPU GFX1310, the RISC-V SpacemiT X100 processor, and an experimental assembler for a number of new RISC-V instructions.

Languages and tools

Clang picked up partial support for the C++26 standard, available via the -std=c++2c flag, along with expanded C2Y functionality. Worth noting separately are the arrival of the x86 Lightweight Fault Isolation (LFI) target — an isolation mechanism for in-process sandboxing — and a new AMD HIP compute-offload driver.

On top of that, the developers improved the LLD linker and worked on the compiler's diagnostic messages, making them clearer.

Where to get it

The source code for LLVM 23.1.0 has already been published on GitHub. Prebuilt binary packages are promised in the near future — for those who would rather not build the compiler themselves.

Source: Phoronix