On Wed, Nov 4, 2020 at 12:03 PM Stephen Hemminger
<step...@networkplumber.org> wrote:
>
> Prefetching only helps performance if it is done several 100
> instructions before the actual use. The purpose of the prefetch
> is to read ahead, it doesn't help if the next instruction
> will block.
>
> The code in the bnxt driver was doing these unnecessary prefetches.
>
> Signed-off-by: Stephen Hemminger <step...@networkplumber.org>
> ---
>  drivers/net/bnxt/bnxt_rxr.c | 4 ----
>  1 file changed, 4 deletions(-)
>
> diff --git a/drivers/net/bnxt/bnxt_rxr.c b/drivers/net/bnxt/bnxt_rxr.c
> index 2645ed61f402..140d79e0e829 100644
> --- a/drivers/net/bnxt/bnxt_rxr.c
> +++ b/drivers/net/bnxt/bnxt_rxr.c
> @@ -305,7 +305,6 @@ static inline struct rte_mbuf *bnxt_tpa_end(
>         mbuf = tpa_info->mbuf;
>         RTE_ASSERT(mbuf != NULL);
>
> -       rte_prefetch0(mbuf);
>         if (agg_bufs) {
>                 bnxt_rx_pages(rxq, mbuf, raw_cp_cons, agg_bufs, tpa_info);
>         }
> @@ -733,8 +732,6 @@ static int bnxt_rx_pkt(struct rte_mbuf **rx_pkt,
>         if (mbuf == NULL)
>                 return -EBUSY;
>
> -       rte_prefetch0(mbuf);
> -
>         mbuf->data_off = RTE_PKTMBUF_HEADROOM;
>         mbuf->nb_segs = 1;
>         mbuf->next = NULL;
> @@ -867,7 +864,6 @@ uint16_t bnxt_recv_pkts(void *rx_queue, struct rte_mbuf 
> **rx_pkts,
>         /* Handle RX burst request */
>         while (1) {
>                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
> -               rte_prefetch0(&cpr->cp_desc_ring[cons]);
>                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
>
>                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
> --
> 2.27.0
>
Acked-by: Lance Richardson <lance.richard...@broadcom.com>

Reply via email to