This fail-safe series should improve some cycles in failsafe data path. It is obvious that we cannot improve a lot because the most of data path functionality are done by the sub devices data path. So the expectation from this series is just to improve some cycles in sub devices management done by fail-safe PMD. I saw ~1.7% MPPS improvement for 1 core 1 queue io-forwarding mode - 2 sub-devices(mlx4 primary).
Matan Azrad (3): net/failsafe: fix Rx safe check compiler hint net/failsafe: mitigate data plan atomic operations net/failsafe: improve Rx sub-devices iteration drivers/net/failsafe/failsafe.c | 5 ++++ drivers/net/failsafe/failsafe_ops.c | 1 + drivers/net/failsafe/failsafe_private.h | 9 +++--- drivers/net/failsafe/failsafe_rxtx.c | 50 ++++++++++----------------------- 4 files changed, 26 insertions(+), 39 deletions(-) -- 1.8.3.1
