Steffen Klassert <steffen.klass...@secunet.com> wrote: > That said, if we want to transmit the 64-bit sequence number > in ESP, I'd prefer to transmit the upper 32-bits before > the lower 32-bits. That's easier on the imlementation side.
My naive notions about cache-line optimizations, I'd think that one could start the replay-window check based upon the lower-32-bits, while waiting for the upper-32-bits to load, in order to validate the upper-32-bits are as expected. For that 1/2^32 situation where the upper-bits roll, then one takes a !unlikely(). I guess the bigger proceedural question is how do we get early feedback from hardware designers. -- Michael Richardson <mcr+i...@sandelman.ca>, Sandelman Software Works -= IPv6 IoT consulting =- *I*LIKE*TRAINS*
signature.asc
Description: PGP signature
_______________________________________________ IPsec mailing list -- ipsec@ietf.org To unsubscribe send an email to ipsec-le...@ietf.org