On Sat, Sep 28, 2024 at 1:51 AM David Marchand
<david.marchand@redhat.com> wrote:
>
> On Fri, Sep 6, 2024 at 1:07 PM Vladimir Medvedkin
> <vladimir.medvedkin@intel.com> wrote:
> >
> > Previously when running rte_fib_lookup IPv4 addresses must have been in
> > host byte order.
> >
> > This patch adds a new flag RTE_FIB_FLAG_LOOKUP_BE that can be passed on
> > fib create, which will allow to have IPv4 in network byte order on
> > lookup.
> >
> > Signed-off-by: Vladimir Medvedkin <vladimir.medvedkin@intel.com>
>
> _mm512_shuffle_epi8 requires avx512bw, so the dir24_8 AVX512
> implementation can only compile when such feature is supported and
> enabled.
> Like the trie AVX512 implem.
>
> IOW, we are missing an update of the library meson.build.

I also suspect that both this added support in the dir24_8
implementation and the existing trie are missing some runtime check on
RTE_CPUFLAG_AVX512BW in get_vector_fn.
Can we get a fix for trie and a respin of this current patch?

Thanks.


--
David Marchand