diff options
| author |   Zhiyong Yang <zhiyong.yang@intel.com> | 2018-11-05 03:45:25 -0500 |
|---|---|---|
| committer |   John Lo <loj@cisco.com> | 2018-11-05 21:02:38 +0000 |
| commit | b0073e276d9e12f02f8f9874fd09ae532a0baa47 (patch) | |
| tree | f2db5ddcf53e9bc48421a6ff82cbee0fcbe75f2d /src/vnet/vxlan | |
| parent | 2130284302788759c7b2644efb801b84411e342e (diff) | |
vxlan: prefetch extra one cacheline holding external header
For vxlan_encap, code will touch memory area before the field "data"
in struct vlib_buffer_t, however so far it is not prefetched in cache
yet for this graph node.
After applying the patch, 2~3 cycles per pkt for vxlan4_encap can be
saved on Haswell. It will bring a lot of benefits on DVN platform too.
Change-Id: I26d8c57fb3d2415726be5367117d73eb715e35ad
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>
Diffstat (limited to 'src/vnet/vxlan')
| -rw-r--r-- | src/vnet/vxlan/encap.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/vnet/vxlan/encap.c b/src/vnet/vxlan/encap.c index fdca425889e..31320900d3e 100644 --- a/src/vnet/vxlan/encap.c +++ b/src/vnet/vxlan/encap.c @@ -116,8 +116,8 @@ vxlan_encap_inline (vlib_main_t * vm, vlib_prefetch_buffer_header (p2, LOAD); vlib_prefetch_buffer_header (p3, LOAD); - CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, LOAD); - CLIB_PREFETCH (p3->data, CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p2->data - CLIB_CACHE_LINE_BYTES, 2 * CLIB_CACHE_LINE_BYTES, LOAD); + CLIB_PREFETCH (p3->data - CLIB_CACHE_LINE_BYTES, 2 * CLIB_CACHE_LINE_BYTES, LOAD); } u32 bi0 = to_next[0] = from[0]; |