> -----Original Message-----
> From: Amber, Kumar <kumar.am...@intel.com>
> Sent: Friday, May 6, 2022 6:23 AM
> To: ovs-dev@openvswitch.org
> Cc: echau...@redhat.com; ktray...@redhat.com; i.maxim...@ovn.org; Ferriter,
> Cian <cian.ferri...@intel.com>; Stokes, Ian <ian.sto...@intel.com>;
> david.march...@redhat.com; f...@sysclose.org; Van Haaren, Harry
> <harry.van.haa...@intel.com>; Amber, Kumar <kumar.am...@intel.com>
> Subject: [PATCH v7 1/4] mfex_avx512: Calculate pkt offsets at compile time.
>
> The patch removes magic numbers pkt offsets and
> minimum packet lenght and instead calculate it at
> compile time.
>
> Signed-off-by: Kumar Amber <kumar.am...@intel.com>
> ---
> lib/dpif-netdev-extract-avx512.c | 28 ++++++++++++++++++++--------
> 1 file changed, 20 insertions(+), 8 deletions(-)
>
> diff --git a/lib/dpif-netdev-extract-avx512.c
> b/lib/dpif-netdev-extract-avx512.c
> index 6b6fe07db..6ae15a4db 100644
> --- a/lib/dpif-netdev-extract-avx512.c
> +++ b/lib/dpif-netdev-extract-avx512.c
> @@ -246,6 +246,16 @@ _mm512_maskz_permutexvar_epi8_wrap(__mmask64
> kmask, __m512i idx, __m512i a)
> NC, NC, NC, NC, 0xBF, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, \
> NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC, NC
>
> +#define PKT_OFFSET_L2 (ETH_HEADER_LEN)
> +#define PKT_OFFSET_L3_VLAN (ETH_HEADER_LEN + VLAN_HEADER_LEN)
VLAN is part of L2, so the naming convention here gets unclear...?
It should be either L2_SIZE or L3_OFFSET.
> +#define PKT_OFFSET_L4_IPv4 (ETH_HEADER_LEN + IP_HEADER_LEN)
as a result of the above, here it looks like IP is L4?
I think PKT_OFFSET_L3_IPV4 would be a clearer name here (as it's the L3
offset).
<snip remaining, will review next version>
> @@ -345,9 +355,9 @@ static const struct mfex_profile
> mfex_profiles[PROFILE_COUNT] =
>
> .mf_bits = { 0x18a0000000000000, 0x0000000000040401},
> .dp_pkt_offs = {
> - 0, UINT16_MAX, 14, 34,
> + 0, UINT16_MAX, PKT_OFFSET_L2, PKT_OFFSET_L4_IPv4,
Usage here is as follows:
l2_pad_size, l2_5_ofs, l3_ofs, l4_ofs
so 0, UINT16_MAX, PKT_L3_OFFSET, PKT_L4_OFFSET would be a nice set of defines
to use.
<snip>
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
