diff options
| author |   Andrew Yourtchenko <ayourtch@gmail.com> | 2018-03-07 17:35:59 +0100 |
|---|---|---|
| committer |   Damjan Marion <dmarion.lists@gmail.com> | 2018-03-11 19:29:27 +0000 |
| commit | 2ca200501e3e58a8f7d2497ff9ba5a39f2f47014 (patch) | |
| tree | 6fcf43ff0f53a62f0c4b6e76167ce10d51b74ad6 /src/vppinfra/test_bihash_vec88.c | |
| parent | eeef247aef46684f2537644b9ef1ade2e313437e (diff) | |
vnet: l2-classify: prefetch (n+2, n+3) rather than (n+1, n+2) inside dual loop code
The dual loop within the L2 classifier processes the (n, n+1) packets,
the prefetching was trying to prefetch (n+1, n+2) - thus half of the
prefetches were not used - because the next iteration needs (n+2, n+3).
Change-Id: I827d20845dbdd0dcdcf463ee25661a7921428992
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
Diffstat (limited to 'src/vppinfra/test_bihash_vec88.c')
0 files changed, 0 insertions, 0 deletions