On May 29, 2014, at 2:09 PM, Ben Pfaff <b...@nicira.com> wrote:
> On Thu, May 29, 2014 at 01:56:31PM -0700, Jarno Rajahalme wrote:
>>
>>> On May 29, 2014, at 1:03 PM, Ben Pfaff <b...@nicira.com> wrote:
>>>
>>> I think that miniflow_extract() goes a little overboard on the
>>> OVS_LIKELYs. In particular these seem suspicious:
>>>
>>> if (OVS_LIKELY((nw_proto != IPPROTO_HOPOPTS)
>>> if (OVS_LIKELY(nw_proto == IPPROTO_TCP)) {
>>> } else if (OVS_LIKELY(nw_proto == IPPROTO_UDP)) {
>>> } else if (OVS_LIKELY(nw_proto == IPPROTO_SCTP)) {
>>> } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMP)) {
>>> } else if (OVS_LIKELY(nw_proto == IPPROTO_ICMPV6)) {
>>>
>>> I would put all of these into a single outermost OVS_LIKELY, to make
>>> it easier to read:
>>> if (OVS_LIKELY(arp) && OVS_LIKELY(arp->ar_hrd == htons(1))
>>> && OVS_LIKELY(arp->ar_pro == htons(ETH_TYPE_IP))
>>> && OVS_LIKELY(arp->ar_hln == ETH_ADDR_LEN)
>>> && OVS_LIKELY(arp->ar_pln == 4)) {
>>>
>>> What do you think?
>>
>> You are LIKELY right, and we can always tune these if there is a
>> practical difference in generated code.
>
> OK.
>
> Your patch or mine?
I don’t have one, so… also as I took it overboard, it might be better you fix
it to your liking, less expected back and forth.
Jarno
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
