Intel Icelake Support Gets Further Into Shape For LLVM Clang 6.0

LLVM's Clang compiler support for the Intel Icelake processors that succeed Cannonlake is getting into better shape ahead of the LLVM/Clang 6.0 feature freeze in January.

Last month -march=icelake was added to LLVM Clang but it mostly mirrored the Cannonlake behavior. Fortunately, now more of the Icelake bits are being filled on the LLVM/Clang side, similar to all the Icelake-related happenings that occurred in the GCC 8 code-base over the past several weeks.

The latest Icelake commits to LLVM/Clang this morning include bitalg (new AVX-512 instruction to Icelake), gfni (Galois Field NI / Galois Field Affine Transformation), vaes (another AVX-512 addition), vpclmulqdq, vnni, and vbmi2.

GCC 8 development code has already supported these instructions while now Clang master does too. LLVM Clang 6.0 should be officially released by early March giving plenty of time for this compiler to work its way to more developer systems with the first Intel Icelake CPUs likely being at least one year out as we currently await the arrival of Cannonlake.