Adhemerval Zanella of Linaro has contributed an FMA-optimized cosh() implementation to glibc, delivering approximately 35% performance improvement over the existing SSE2 version when targeting the x86_64-v3 micro-architecture level. Additional patches merged alongside this change also improve tanh and sinh performance using CORE-MATH, though results for those are mixed depending on target CPU/ISA. This work is building toward the Glibc 2.44 release scheduled for August.
Sort: