Hi Bruce, On 08/11/2014 10:44 PM, Bruce Richardson wrote: > Adjust the fast-path code to fix the regression caused by the pool > pointer moving to the second cache line. This change adjusts the > prefetching and also the way in which the mbufs are freed back to the > mempool. > > Signed-off-by: Bruce Richardson <bruce.richardson at intel.com>
Just one comment here (maybe this code should be reviewed by someone knowing the ixgbe driver better than me): > @@ -252,14 +250,6 @@ tx_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, > volatile union ixgbe_adv_tx_desc *tx_r = txq->tx_ring; > uint16_t n = 0; > > - /* > - * Begin scanning the H/W ring for done descriptors when the > - * number of available descriptors drops below tx_free_thresh. For > - * each done descriptor, free the associated buffer. > - */ > - if (txq->nb_tx_free < txq->tx_free_thresh) > - ixgbe_tx_free_bufs(txq); > - > /* Only use descriptors that are available */ > nb_pkts = (uint16_t)RTE_MIN(txq->nb_tx_free, nb_pkts); > if (unlikely(nb_pkts == 0)) > @@ -323,6 +313,15 @@ tx_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, > if (txq->tx_tail >= txq->nb_tx_desc) > txq->tx_tail = 0; > > + /* > + * Begin scanning the H/W ring for done descriptors when the > + * number of available descriptors drops below tx_free_thresh. For > + * each done descriptor, free the associated buffer. > + */ > + if (txq->nb_tx_free < txq->tx_free_thresh) > + ixgbe_tx_free_bufs(txq); > + > + > /* update tail pointer */ > rte_wmb(); > IXGBE_PCI_REG_WRITE(txq->tdt_reg_addr, txq->tx_tail); It looks like these 2 hunks are reverted in next commit. I'm not sure this is what you expected.
