Hi Eelco,
Replies inline.
<Snip>
> > + * The functionality here can be easily auto-validated and tested
> > +against the
> > + * scalar miniflow_extract() function. As such, manual review of the
> > +code by
> > + * the community (although welcome) is not required. Confidence in
> > +the
> > + * correctness of the code can be had from the autovalidation.
>
> The last sentence sounds odd, “can be had from the autovalidation?”.
Fixed the line.
>
> > + */
> > +static const struct mfex_profile mfex_profiles[PROFILE_COUNT] = {
> > + [PROFILE_ETH_IPV4_UDP] = {
> > + .probe_mask.u8_data = { PATTERN_ETHERTYPE_MASK
> PATTERN_IPV4_MASK },
> > + .probe_data.u8_data = { PATTERN_ETHERTYPE_IPV4
> > +PATTERN_IPV4_UDP},
> > +
> > + .store_shuf.u8_data = { PATTERN_IPV4_UDP_SHUFFLE },
> > + .store_kmsk = PATTERN_IPV4_UDP_KMASK,
> > +
> > + .mf_bits = { 0x18a0000000000000, 0x0000000000040401},
>
> As discussed in the v4 review, please update the comments text above the flow
> structure definition to include the AVX functions to be updated?
>
> 97 * NOTE: Order of the fields is significant, any change in the order must
> be
> 98 * reflected in miniflow_extract()!
>
Added in v7.
> > + .dp_pkt_offs = {
> > + 0, UINT16_MAX, 14, 34,
> > + },
> > + .dp_pkt_min_size = 42,
> > - for (int i = MFEX_IMPL_MAX; i < impl_count; i++) {
> > + for (int i = MFEX_IMPL_VMBI_IPv4_UDP; i < impl_count; i++) {
>
> Guess all instance of MFEX_IMPL_VMBI_IPv4_UDP need to be replaced with,
> MFEX_IMPL_START_IDX, as confirmed earlier.
>
Already fixed .
> > if (miniflow_funcs[i].available) {
> > hitmask = miniflow_funcs[i].extract_func(packets, keys,
> > keys_size,
Regards
Amber
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
