diff options
| author |   PiotrX Kleski <piotrx.kleski@intel.com> | 2020-12-09 14:32:26 +0100 |
|---|---|---|
| committer |   Damjan Marion <dmarion@me.com> | 2020-12-11 23:29:59 +0000 |
| commit | b801cd1b31e44e9e90f4619db0cad4ab661019e5 (patch) | |
| tree | 6fa8fc4a54fef7493cbffdc023540fb633dd8206 /src | |
| parent | 32526a42c59f076e356c3690306b8bcaa80d999d (diff) | |
ip: ip4 rewrite prefetch optimization
Type: improvement
ip4_rewrite_inline_with_gso() did vlib_prefetch_buffer_header() for all nodes.
However it is not necessary for ip-rewrite, it is only needed by ip-midchain.
This patch makes ip4-rewrite prefetches less buffers to save cycles.
Signed-off-by: PiotrX Kleski <piotrx.kleski@intel.com>
Change-Id: Ib82dcb0eda4a2d1d7b8d664f2224d49b72aef50f
Diffstat (limited to 'src')
| -rw-r--r-- | src/vnet/ip/ip4_forward.c | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/src/vnet/ip/ip4_forward.c b/src/vnet/ip/ip4_forward.c index fdc92ca16f8..77ea4c5524e 100644 --- a/src/vnet/ip/ip4_forward.c +++ b/src/vnet/ip/ip4_forward.c @@ -2109,8 +2109,11 @@ ip4_rewrite_inline_with_gso (vlib_main_t * vm, u32 tx_sw_if_index0, tx_sw_if_index1; u8 *p; - vlib_prefetch_buffer_header (b[6], LOAD); - vlib_prefetch_buffer_header (b[7], LOAD); + if (is_midchain) + { + vlib_prefetch_buffer_header (b[6], LOAD); + vlib_prefetch_buffer_header (b[7], LOAD); + } adj_index0 = vnet_buffer (b[0])->ip.adj_index[VLIB_TX]; adj_index1 = vnet_buffer (b[1])->ip.adj_index[VLIB_TX]; |