P3 Description
- AVX-512 masked move instructions can be used to generate efficient code sequence for tails or small copy sizes.
- Through partial inlining, we can prevent making a call to stubs thus save on call overhead for small copy sizes.
- Optimize array copy stubs through aligned loops to prevent cache line split penalty for large copy sizes.
- Through partial inlining, we can prevent making a call to stubs thus save on call overhead for small copy sizes.
- Optimize array copy stubs through aligned loops to prevent cache line split penalty for large copy sizes.