AMD AOCC 6.0 Brings Zen 6 Optimizations, New "aof" Fortran Compiler, aoc / aoc++ Branding

With yesterday's launch availability of the AMD EPYC 9006 "Venice" SP7 processors headlined by the incredibly fast EPYC 9996 and EPYC 9686F processors, AMD also launched AOCC 6.0 as the newest version of their in-house LLVM/Clang-based compiler stack. AOCC 6.0 brings their initial round of AMD Zen 6 tuning catering to EPYC Venice processors and an alternative to the existing "znver6" target support found in upstream LLVM/Clang as well as the GCC compiler. AOCC 6.0 brings AOCL-LibM 6.0 that is tuned for Zen 6, the AOCC code at large is tuned for EPYC 9006 series, and other updates. AOCC 6.0 also introduces aof as their new Fortran compiler. AOF is the future of AMD's AOCC Fortran efforts and currently on-par with the upstream LLVM Clang compiler code. This new compiler offers better conformance to the latest Fortran and OpenMP standards. The modified Flang compiler is still in the AOCC codebase with the v6.0 release. With AOCC releases up to now the compiler binaries have been clang and clang++ while now they are branded to better distinguish the AOCC compiler use. There is now aoc as the name of AMD's C compiler and aoc++ as the AMD C++ compiler. The clang and clang++ are symlinked to these new branded binaries. AOCC 6.0 is derived from upstream LLVM 21.1.8. With the LLVM 21 series being from the end of last year and LLVM 23 being the latest stable series, it's a bit unfortunate they weren't able to pull in LLVM 22 or LLVM 23 as the basis for their updated compiler. Downloads and more details on the AMD Optimizing C/C++ Compiler 6.0 release can be found via AMD.com. As time allows on my AMD EPYC 9006 SP7 test server I will be putting the new AOCC 6.0 compiler release through some rigorous benchmarks to see how it compares to upstream LLVM/Clang and GCC.