> From: Stephen Hemminger [mailto:step...@networkplumber.org] > Sent: Saturday, 2 March 2024 21.49 > > The rte_memcpy code would do extra instructions for size 16 > and 32 which potentially could reference past end of data.
It's a somewhat weird concept, but they don't reference past end of data. They reference data in chunks of 16. E.g. when copying 17 bytes: first copy [0..15] and then copy [1..16] (as "-16 + n" in the code). By referencing an address "-16" in a block of 16 bytes, they are not referencing past end of data. > > For size of 16, only single mov16 is needed. > same for size of 32, only single mov32. I fixed the duplicate copies with my patch [1]. Please review. [1]: https://inbox.dpdk.org/dev/20240302234812.9137-1...@smartsharesystems.com/T/#u
