On Fri, Aug 27, 2021 at 3:12 PM Ganapati Kundapura
<ganapati.kundap...@intel.com> wrote:
>
> RX adapter user memove() to move unprocessed events to the beginning of
> the packet enqueue buffer. The use memmove() was found to consume good
> amount of CPU cycles (about 20%).
>
> This patch removes the use of memove() while implementina a circular
Typo
> buffer to avoid copying of data. With this change RX adapter is able
> to fill the buffer of 16384 events.
Please change the subject to :
eventdev: rx-adapter: improve ...
>
> Signed-off-by: Ganapati Kundapura <ganapati.kundap...@intel.com>
> ---
> lib/eventdev/rte_event_eth_rx_adapter.c | 84
> ++++++++++++++++++++++++++-------
> 1 file changed, 68 insertions(+), 16 deletions(-)
>
> } else {
> num = rxa_create_event_vector(rx_adapter, eth_rx_queue_info,
> @@ -892,9 +918,14 @@ rxa_buffer_mbufs(struct rte_event_eth_rx_adapter
> *rx_adapter,
>
> dropped = 0;
> nb_cb = dev_info->cb_fn(eth_dev_id, rx_queue_id,
> - ETH_EVENT_BUFFER_SIZE, buf->count,
> - &buf->events[buf->count], num,
> - dev_info->cb_arg, &dropped);
> + buf->last |
> + (RTE_DIM(buf->events) &
> ~buf->last_mask),
> + buf->count >= BATCH_SIZE ?
> + buf->count - BATCH_SIZE : 0,
> + &buf->events[buf->tail],
> + num,
> + dev_info->cb_arg,
> + &dropped);
Moving this code section to a separate inline function can allow more
linewidth to consume aka
less number of lines.
Rest looks good to me.
