Arm Neoverse V1 Support Added To LLVM Clang 12 Compiler

Announced by Arm back in September was the Neoverse V1 "Zeus" platform with 50%+ performance uplift over the Neoverse N1. Neoverse V1 also supports Scalable Vector Extensions (SVE) and other features in catering the platform for machine learning, cloud, HPC, and similar use-cases.

Quite quickly after that GNU toolchain support began appearing while now the LLVM Clang compiler support has their Neoverse V1 target merged thanks to the punctual work of Arm engineers.

Beginning with the LLVM Clang 12 release due out next February~March, the "neoverse-v1" target is now supported as a valid -march/mcpu option.

Specifying "neoverse-v1" for the compiler implies ARMv8.4-A plus features of BF16, cache deep persist, hardware crypto, FP16FML, full FP16, fuse AES, matrix multiply INT8, NEON, PerfMon, RandGen, SPE, SSBS, and SVE instructions.

The initial Arm Neoverse V1 support was merged this morning into the LLVM 12 Git repository.