net/mlx5: Use 128B cacheline size for 128B or larger cachelines

The adapter uses the cache_line_128byte setting to set the bounds for
end padding. On systems where the cacheline size is greater than 128B
use 128B instead of the default of 64B. This results in fewer partial
cacheline writes. There's a 50% chance it will pad to the end of a 256B
cache line vs only 25% when using 64B.

Fixes: f32f5bd2eb ("net/mlx5: Configure cache line size for start and end padding")
Signed-off-by: Daniel Jurgens <danielj@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This commit is contained in:
Daniel Jurgens 2018-02-02 09:32:53 -06:00 committed by Saeed Mahameed
parent 2f0db87901
commit c67f100eda
1 changed files with 1 additions and 1 deletions

View File

@ -551,7 +551,7 @@ static int handle_hca_cap(struct mlx5_core_dev *dev)
MLX5_SET(cmd_hca_cap,
set_hca_cap,
cache_line_128byte,
cache_line_size() == 128 ? 1 : 0);
cache_line_size() >= 128 ? 1 : 0);
if (MLX5_CAP_GEN_MAX(dev, dct))
MLX5_SET(cmd_hca_cap, set_hca_cap, dct, 1);