Message ID | 20110328135041.57cec646@kryten |
---|---|
State | Accepted, archived |
Delegated to: | David Miller |
Headers | show |
On 03/27/2011 07:50 PM, Anton Blanchard wrote: > > > While testing the performance of different receive interrupt > coalescing settings on a single stream TCP benchmark, I noticed two > very different results. With rx-usecs=50, most of the time a > connection would hit 8280 Mbps but once in a while it would hit > 9330 Mbps. > > It turns out we are only applying the interrupt coalescing settings > to the first queue and whenever the rx hash would direct us onto > that queue we ran faster. > > With this patch applied and rx-usecs=50, I get 9330 Mbps > consistently. > > Signed-off-by: Anton Blanchard <anton@samba.org> Acked-by: Divy Le Ray <divy@chelsio.com> > --- > > Index: powerpc.git/drivers/net/cxgb3/cxgb3_main.c > =================================================================== > --- powerpc.git.orig/drivers/net/cxgb3/cxgb3_main.c 2011-03-28 > 12:36:41.775206214 +1100 > +++ powerpc.git/drivers/net/cxgb3/cxgb3_main.c 2011-03-28 > 12:41:29.892189120 +1100 > @@ -1983,14 +1983,20 @@ static int set_coalesce(struct net_devic > { > struct port_info *pi = netdev_priv(dev); > struct adapter *adapter = pi->adapter; > - struct qset_params *qsp = &adapter->params.sge.qset[0]; > - struct sge_qset *qs = &adapter->sge.qs[0]; > + struct qset_params *qsp; > + struct sge_qset *qs; > + int i; > > if (c->rx_coalesce_usecs * 10 > M_NEWTIMER) > return -EINVAL; > > - qsp->coalesce_usecs = c->rx_coalesce_usecs; > - t3_update_qset_coalesce(qs, qsp); > + for (i = 0; i < pi->nqsets; i++) { > + qsp = &adapter->params.sge.qset[i]; > + qs = &adapter->sge.qs[i]; > + qsp->coalesce_usecs = c->rx_coalesce_usecs; > + t3_update_qset_coalesce(qs, qsp); > + } > + > return 0; > } > > -- > To unsubscribe from this list: send the line "unsubscribe netdev" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Divy Le Ray <divy@chelsio.com> Date: Mon, 28 Mar 2011 10:41:54 -0700 > On 03/27/2011 07:50 PM, Anton Blanchard wrote: >> >> >> While testing the performance of different receive interrupt >> coalescing settings on a single stream TCP benchmark, I noticed two >> very different results. With rx-usecs=50, most of the time a >> connection would hit 8280 Mbps but once in a while it would hit >> 9330 Mbps. >> >> It turns out we are only applying the interrupt coalescing settings >> to the first queue and whenever the rx hash would direct us onto >> that queue we ran faster. >> >> With this patch applied and rx-usecs=50, I get 9330 Mbps >> consistently. >> >> Signed-off-by: Anton Blanchard <anton@samba.org> > > Acked-by: Divy Le Ray <divy@chelsio.com> Applied, thanks. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Index: powerpc.git/drivers/net/cxgb3/cxgb3_main.c =================================================================== --- powerpc.git.orig/drivers/net/cxgb3/cxgb3_main.c 2011-03-28 12:36:41.775206214 +1100 +++ powerpc.git/drivers/net/cxgb3/cxgb3_main.c 2011-03-28 12:41:29.892189120 +1100 @@ -1983,14 +1983,20 @@ static int set_coalesce(struct net_devic { struct port_info *pi = netdev_priv(dev); struct adapter *adapter = pi->adapter; - struct qset_params *qsp = &adapter->params.sge.qset[0]; - struct sge_qset *qs = &adapter->sge.qs[0]; + struct qset_params *qsp; + struct sge_qset *qs; + int i; if (c->rx_coalesce_usecs * 10 > M_NEWTIMER) return -EINVAL; - qsp->coalesce_usecs = c->rx_coalesce_usecs; - t3_update_qset_coalesce(qs, qsp); + for (i = 0; i < pi->nqsets; i++) { + qsp = &adapter->params.sge.qset[i]; + qs = &adapter->sge.qs[i]; + qsp->coalesce_usecs = c->rx_coalesce_usecs; + t3_update_qset_coalesce(qs, qsp); + } + return 0; }
While testing the performance of different receive interrupt coalescing settings on a single stream TCP benchmark, I noticed two very different results. With rx-usecs=50, most of the time a connection would hit 8280 Mbps but once in a while it would hit 9330 Mbps. It turns out we are only applying the interrupt coalescing settings to the first queue and whenever the rx hash would direct us onto that queue we ran faster. With this patch applied and rx-usecs=50, I get 9330 Mbps consistently. Signed-off-by: Anton Blanchard <anton@samba.org> --- -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html