From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] pktgen node allocation Date: Fri, 19 Mar 2010 14:47:45 +0100 Message-ID: <1269006465.3048.39.camel@edumazet-laptop> References: <19363.14702.909265.380669@gargle.gargle.HOWL> <1268990933.3048.15.camel@edumazet-laptop> <19363.32154.39665.185451@gargle.gargle.HOWL> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , netdev@vger.kernel.org To: robert@herjulf.net Return-path: Received: from mail-bw0-f209.google.com ([209.85.218.209]:37232 "EHLO mail-bw0-f209.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752256Ab0CSNru (ORCPT ); Fri, 19 Mar 2010 09:47:50 -0400 Received: by bwz1 with SMTP id 1so42355bwz.21 for ; Fri, 19 Mar 2010 06:47:49 -0700 (PDT) In-Reply-To: <19363.32154.39665.185451@gargle.gargle.HOWL> Sender: netdev-owner@vger.kernel.org List-ID: Le vendredi 19 mars 2010 =C3=A0 14:35 +0100, robert@herjulf.net a =C3=A9= crit : > Eric Dumazet writes: > > Le vendredi 19 mars 2010 =C3=A0 09:44 +0100, Robert Olsson a =C3=A9= crit : > >=20 > > I cannot understand how this can help. > >=20 > > __netdev_alloc_skb() is supposed to already take into account NUMA > > properties : > >=20 > > int node =3D dev->dev.parent ? dev_to_node(dev->dev.parent) : -1; > >=20 > > If this doesnt work, we should correct core stack, not only pktgen= :) > >=20 > > Are you allocating memory in the node where pktgen CPU is running = or the > > node close to the NIC ? >=20 > I didn't say it should help the idea was to give some hooks to=20 > experiment and see effects with different node memory allocations. > There are many degrees of freedom wrt buses(device)/CPU/menory. >=20 Well, you said "Tested this with 10 Intel 82599 ports w. TYAN S7025 E5520 CPU's. Was able to TX/DMA ~80 Gbit/s to Ethernet wires." I am interested to know what particular setup you did to maximize throughput then, or are you saing you managed to reduce it ? :)