[PATCH v2] net/i40e: remove redundant judgment
Zhang, Qi Z
qi.z.zhang at intel.com
Thu Apr 27 09:37:53 CEST 2023
> -----Original Message-----
> From: Feifei Wang <feifei.wang2 at arm.com>
> Sent: Tuesday, March 28, 2023 3:28 PM
> To: Richardson, Bruce <bruce.richardson at intel.com>; Konstantin Ananyev
> <konstantin.v.ananyev at yandex.ru>; Zhang, Yuying
> <yuying.zhang at intel.com>; Xing, Beilei <beilei.xing at intel.com>; David
> Christensen <drc at linux.vnet.ibm.com>; Ruifeng Wang
> <ruifeng.wang at arm.com>
> Cc: dev at dpdk.org; nd at arm.com; Feifei Wang <feifei.wang2 at arm.com>;
> Honnappa Nagarahalli <honnappa.nagarahalli at arm.com>
> Subject: [PATCH v2] net/i40e: remove redundant judgment
>
> Merged variable updates under the same condition. It reduces branch.
>
> In ampere-altra, there is no performance improvement with this patch.
> In x86 sse and avx2 path, there is also no performance improvement.
Thanks for sharing the results. While the code implements some best practices, such as reducing branching and adding compiler hints, which should generally improve performance, it's not necessary to highlight that it didn't provide benefits on certain specific platforms.
Would it be ok to remove the last two lines when merging the patch?
Otherwise
Acked-by: Qi Zhang <qi.z.zhang at intel.com>
>
>
> v2:
> 1. add change for avx and altivec path.
>
> Suggested-by: Honnappa Nagarahalli <honnappa.nagarahalli at arm.com>
> Signed-off-by: Feifei Wang <feifei.wang2 at arm.com>
> Reviewed-by: Ruifeng Wang <ruifeng.wang at arm.com>
> ---
> drivers/net/i40e/i40e_rxtx_common_avx.h | 9 +++++----
> drivers/net/i40e/i40e_rxtx_vec_altivec.c | 9 +++++----
> drivers/net/i40e/i40e_rxtx_vec_neon.c | 9 +++++----
> drivers/net/i40e/i40e_rxtx_vec_sse.c | 9 +++++----
> 4 files changed, 20 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/net/i40e/i40e_rxtx_common_avx.h
> b/drivers/net/i40e/i40e_rxtx_common_avx.h
> index cfc1e63173..85958d6c81 100644
> --- a/drivers/net/i40e/i40e_rxtx_common_avx.h
> +++ b/drivers/net/i40e/i40e_rxtx_common_avx.h
> @@ -198,14 +198,15 @@ i40e_rxq_rearm_common(struct i40e_rx_queue
> *rxq, __rte_unused bool avx512) #endif
>
> rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
> - if (rxq->rxrearm_start >= rxq->nb_rx_desc)
> + rx_id = rxq->rxrearm_start - 1;
> +
> + if (unlikely(rxq->rxrearm_start >= rxq->nb_rx_desc)) {
> rxq->rxrearm_start = 0;
> + rx_id = rxq->nb_rx_desc - 1;
> + }
>
> rxq->rxrearm_nb -= RTE_I40E_RXQ_REARM_THRESH;
>
> - rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
> - (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
> -
> /* Update the tail pointer on the NIC */
> I40E_PCI_REG_WC_WRITE(rxq->qrx_tail, rx_id); } diff --git
> a/drivers/net/i40e/i40e_rxtx_vec_altivec.c
> b/drivers/net/i40e/i40e_rxtx_vec_altivec.c
> index 2dfa04599c..8672ad1c41 100644
> --- a/drivers/net/i40e/i40e_rxtx_vec_altivec.c
> +++ b/drivers/net/i40e/i40e_rxtx_vec_altivec.c
> @@ -89,14 +89,15 @@ i40e_rxq_rearm(struct i40e_rx_queue *rxq)
> }
>
> rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
> - if (rxq->rxrearm_start >= rxq->nb_rx_desc)
> + rx_id = rxq->rxrearm_start - 1;
> +
> + if (unlikely(rxq->rxrearm_start >= rxq->nb_rx_desc)) {
> rxq->rxrearm_start = 0;
> + rx_id = rxq->nb_rx_desc - 1;
> + }
>
> rxq->rxrearm_nb -= RTE_I40E_RXQ_REARM_THRESH;
>
> - rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
> - (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
> -
> /* Update the tail pointer on the NIC */
> I40E_PCI_REG_WRITE(rxq->qrx_tail, rx_id); } diff --git
> a/drivers/net/i40e/i40e_rxtx_vec_neon.c
> b/drivers/net/i40e/i40e_rxtx_vec_neon.c
> index 12e6f1cbcb..49391fe4c7 100644
> --- a/drivers/net/i40e/i40e_rxtx_vec_neon.c
> +++ b/drivers/net/i40e/i40e_rxtx_vec_neon.c
> @@ -64,14 +64,15 @@ i40e_rxq_rearm(struct i40e_rx_queue *rxq)
> }
>
> rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
> - if (rxq->rxrearm_start >= rxq->nb_rx_desc)
> + rx_id = rxq->rxrearm_start - 1;
> +
> + if (unlikely(rxq->rxrearm_start >= rxq->nb_rx_desc)) {
> rxq->rxrearm_start = 0;
> + rx_id = rxq->nb_rx_desc - 1;
> + }
>
> rxq->rxrearm_nb -= RTE_I40E_RXQ_REARM_THRESH;
>
> - rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
> - (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
> -
> rte_io_wmb();
> /* Update the tail pointer on the NIC */
> I40E_PCI_REG_WRITE_RELAXED(rxq->qrx_tail, rx_id); diff --git
> a/drivers/net/i40e/i40e_rxtx_vec_sse.c
> b/drivers/net/i40e/i40e_rxtx_vec_sse.c
> index bdc979a839..baf83cb3df 100644
> --- a/drivers/net/i40e/i40e_rxtx_vec_sse.c
> +++ b/drivers/net/i40e/i40e_rxtx_vec_sse.c
> @@ -77,14 +77,15 @@ i40e_rxq_rearm(struct i40e_rx_queue *rxq)
> }
>
> rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
> - if (rxq->rxrearm_start >= rxq->nb_rx_desc)
> + rx_id = rxq->rxrearm_start - 1;
> +
> + if (unlikely(rxq->rxrearm_start >= rxq->nb_rx_desc)) {
> rxq->rxrearm_start = 0;
> + rx_id = rxq->nb_rx_desc - 1;
> + }
>
> rxq->rxrearm_nb -= RTE_I40E_RXQ_REARM_THRESH;
>
> - rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
> - (rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
> -
> /* Update the tail pointer on the NIC */
> I40E_PCI_REG_WC_WRITE(rxq->qrx_tail, rx_id); }
> --
> 2.25.1
More information about the dev
mailing list