AMD has published a compiler (AMD Optimizing C/C++ Compiler) built on top of LLVM and including additional enhancements and optimizations for 17 families of AMD processors based on microarchitectures , ΠΈ , for example, for already released AMD Ryzen and EPYC processors. The compiler also contains general improvements related to vectorization, code generation, high-level optimization, inter-procedural analysis, and loop transformation. By default, the LLD linker is enabled. The package includes an optimized version of the libm mathematical library - AMDLibM. The compiler is available for 32-bit and 64-bit Linux systems.
In the new release, the codebase has been updated to branch . Added optimizations for the AMD EPYC 7002 Series (Zen 2) architecture, which has improved code generation and vectorization. To enable optimizations for Zen 2, the option to select the "znver2" architecture is proposed. Support for the Flang compiler for the Fortran language has been provided. The AMDLibM library has been updated to release 3.3. The binaries provided for download have been tested with RHEL 7.4, SLES 12 SP3, and Ubuntu 18.04 LTS. AOCC is currently distributed only in binary form and requires the acceptance of the EULA agreement.
Source: opennet.ru