Fujitsu A64FX Support Added To The LLVM Clang 11 Compiler

Fujitsu has contributed support for their high-performance A64FX ARMv8-based CPU cores to the LLVM Clang compiler.

The Fujitsu A64FX is ARMv8.2-based with 512-bit SIMD via the Arm Scalable Vector Extensions (SVE), 48 cores, and has onboard 32GB of HBM2 memory. We should be seeing more servers / HPC systems coming to market this year with the A64FX processors. Among the announced users of the Fujitsu A64FX is Japan's Post-K exascale supercomputer.

Fujitsu has been providing some open-source support around the A64FX while beginning with LLVM Clang 11 later this year will be mainline toolchain support.

A Fujitsu engineer has contributed the A64FX tuning support to this open-source compiler. The -mcpu=a64fx target is added for this Arm HPC processor to turn on the SVE/SVE2 and FP16 usage. The A64FX support isn't yet complete but at least one area not yet implemented by today's patches is the scheduling model. Nevertheless, it's good seeing Fujitsu being punctual with upstreaming their enablement code.