Usage Description

The vector implementation of servers in the Arm architecture is in NEON or SVE instructions. The AVX2KI library includes the AVX2NEON library, AVX2SVE library, and AVX2SVE512 library. AVX2NEON is for servers in the Armv8 architecture, and AVX2SVE and AVX2SVE512 are for servers in the Armv9 architecture. When KSL is being installed, the AVX2KI library that best adapts to the current hardware is automatically created. You can directly link the AVX2KI library based on 3.

• To use the AVX2SVE dynamic library, add the compilation option -msve-vector-bits=256 to the application layer. To use the AVX2SVE512 dynamic library, add the compilation option -msve-vector-bits=512 to the application layer. These compilation options are supported only in GCC 10 and later versions.
• When applying the AVX2SVE dynamic library, if you need to use SVE instruction types in unions such as __m256i and __m256d, add the compilation option -DTARGET_AVX2SVE to the application layer to make the macro take effect. When applying the AVX2SVE512 dynamic library, add the compilation option -DTARGET_AVX2SVE512 to the application layer.