On 2024/9/20 3:05, Stephen Hemminger wrote: > On Wed, 18 Sep 2024 14:09:36 +0800 > "WanRenyong" <wa...@yunsilicon.com> wrote: > >> @@ -166,6 +169,7 @@ xsc_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, >> uint16_t pkts_n) >> /* Fill wqe */ >> wqe->va = rte_cpu_to_le_64(rte_pktmbuf_iova(rep)); >> rte_pktmbuf_data_len(seg) = len; >> + rxq->stats.rx_bytes += rte_pktmbuf_pkt_len(pkt); >> >> *(pkts++) = pkt; >> pkt = NULL; >> @@ -200,6 +204,7 @@ xsc_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, >> uint16_t pkts_n) >> rxq->nb_rx_hold = 0; >> } >> >> + rxq->stats.rx_pkts += nb_pkts; >> return nb_pkts; >> } > You can a small performance boost by keeping a total_bytes counter in > rx_burst function > and only add it to rx_bytes after the loop. Hello, Stephen,
Thanks for review. This a good suggestion, I'll take it. -- Thanks, WanRenyong