On Mon, 14 Nov 2022 11:41:06 +0100 Pablo Neira Ayuso wrote: > From: Florian Westphal <fw@xxxxxxxxx> > > This function is used for every packet, siphash_4u64 is noticeably faster > than using local buffer + siphash: > > Before: > 1.23% kpktgend_0 [kernel.vmlinux] [k] __siphash_unaligned > 0.14% kpktgend_0 [nf_conntrack] [k] hash_conntrack_raw > After: > 0.79% kpktgend_0 [kernel.vmlinux] [k] siphash_4u64 > 0.15% kpktgend_0 [nf_conntrack] [k] hash_conntrack_raw > > In the pktgen test this gives about ~2.4% performance improvement. Hi, Mr Nit Pick here, can we silence the spatse warnings? net/netfilter/nf_conntrack_core.c:222:14: warning: cast from restricted __be16 net/netfilter/nf_conntrack_core.c:222:55: warning: restricted __be16 degrades to integer
