[PATCH v2 1/2] net/i40e: replace put function

[Morten Brørup]

> From: Kamalakshitha Aligeri [mailto:kamalakshitha.aligeri at arm.com]
> Sent: Friday, 10 February 2023 07.54
> 
> Integrated zero-copy put API in mempool cache in i40e PMD.
> On Ampere Altra server, l3fwd single core's performance improves by 5%
> with the new API
> 
> Signed-off-by: Kamalakshitha Aligeri <kamalakshitha.aligeri at arm.com>
> Reviewed-by: Ruifeng Wang <ruifeng.wang at arm.com>
> Reviewed-by: Feifei Wang <feifei.wang2 at arm.com>
> ---
> Link:
> https://patchwork.dpdk.org/project/dpdk/patch/20230209145833.129986-1-
> mb at smartsharesystems.com/

If you agree with the referred patch, please review or acknowledge it on the mailing list, so it can be merged.

> 
>  .mailmap                                |  1 +
>  drivers/net/i40e/i40e_rxtx_vec_common.h | 28 ++++++++++++++++++++-----
>  2 files changed, 24 insertions(+), 5 deletions(-)
> 
> diff --git a/.mailmap b/.mailmap
> index 75884b6fe2..05a42edbcf 100644
> --- a/.mailmap
> +++ b/.mailmap
> @@ -670,6 +670,7 @@ Kai Ji <kai.ji at intel.com>
>  Kaiwen Deng <kaiwenx.deng at intel.com>
>  Kalesh AP <kalesh-anakkur.purayil at broadcom.com>
>  Kamalakannan R <kamalakannan.r at intel.com>
> +Kamalakshitha Aligeri <kamalakshitha.aligeri at arm.com>
>  Kamil Bednarczyk <kamil.bednarczyk at intel.com>
>  Kamil Chalupnik <kamilx.chalupnik at intel.com>
>  Kamil Rytarowski <kamil.rytarowski at caviumnetworks.com>
> diff --git a/drivers/net/i40e/i40e_rxtx_vec_common.h
> b/drivers/net/i40e/i40e_rxtx_vec_common.h
> index fe1a6ec75e..113599d82b 100644
> --- a/drivers/net/i40e/i40e_rxtx_vec_common.h
> +++ b/drivers/net/i40e/i40e_rxtx_vec_common.h
> @@ -95,18 +95,36 @@ i40e_tx_free_bufs(struct i40e_tx_queue *txq)
> 
>  	n = txq->tx_rs_thresh;
> 
> -	 /* first buffer to free from S/W ring is at index
> -	  * tx_next_dd - (tx_rs_thresh-1)
> -	  */
> +	/* first buffer to free from S/W ring is at index
> +	 * tx_next_dd - (tx_rs_thresh-1)
> +	 */
>  	txep = &txq->sw_ring[txq->tx_next_dd - (n - 1)];
> 
>  	if (txq->offloads & RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) {
> +		struct rte_mempool *mp = txep[0].mbuf->pool;
> +		struct rte_mempool_cache *cache =
> rte_mempool_default_cache(mp, rte_lcore_id());
> +		void **cache_objs;
> +
> +		if (unlikely(!cache))
> +			goto fallback;
> +
> +		cache_objs = rte_mempool_cache_zc_put_bulk(cache, mp, n);
> +		if (unlikely(!cache_objs))
> +			goto fallback;
> +
>  		for (i = 0; i < n; i++) {
> -			free[i] = txep[i].mbuf;
> +			cache_objs[i] = txep->mbuf;
>  			/* no need to reset txep[i].mbuf in vector path */
> +			txep++;

Why the change from "xyz[i] = txep[i].mbuf;" to "xyz[i] = txep->mbuf; txep++;"? I don't see "txep" being used after the "done" label. And at the fallback, you still use "xyz[i] = txep[i].mbuf;". It would look cleaner using the same method in both places.

It's not important, so feel free to keep as is or change as suggested. Both ways,

Acked-by: Morten Brørup <mb at smartsharesystems.com>

>  		}
> -		rte_mempool_put_bulk(free[0]->pool, (void **)free, n);
>  		goto done;
> +
> +fallback:
> +		for (i = 0; i < n; i++)
> +			free[i] = txep[i].mbuf;
> +		rte_mempool_generic_put(mp, (void **)free, n, cache);
> +		goto done;
> +
>  	}
> 
>  	m = rte_pktmbuf_prefree_seg(txep[0].mbuf);
> --
> 2.25.1
>