LLVM 23.1 is a substantial compiler update focused on next-generation CPU/GPU hardware, newer C/C++ standards, and improved security/offloading. (Phoronix)

Key highlights:

• New hardware support: AMD Zen 6 (znver6) with AVX-512 BMM, AMD GFX1310/RDNA 5 GPUs, NVIDIA Rigel, Arm C1-Ultra, and SpacemiT X100 RISC-V.
• x86-64 LFI: Adds Lightweight Fault Isolation targeting, intended to sandbox native C/C++/assembly code within the same process with relatively low overhead. (LLVM)
• C/C++ improvements: Partial C++26 support via -std=c++2c plus additional upcoming C2Y features.
• AMD GPU computing: Introduces the newer HIP offload driver, improving LLVM/Clang’s GPU-offloading infrastructure.
• RISC-V: Adds experimental assembler support for several newer ISA extensions.
• Toolchain improvements: Better LLD linker, diagnostics, and general compiler behavior.

In short: LLVM 23.1 prepares Clang/LLVM for upcoming Zen 6, RDNA 5, NVIDIA Rosa/Rigel, newer Arm and RISC-V hardware, while advancing C++26/C2Y, GPU compute, and native-code sandboxing. (Phoronix)

https://github.com/llvm/llvm-project/releases/tag/llvmorg-23.1.0