From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id E7FD65A13 for ; Thu, 24 Dec 2015 02:33:46 +0100 (CET) Received: from orsmga003.jf.intel.com ([10.7.209.27]) by fmsmga101.fm.intel.com with ESMTP; 23 Dec 2015 17:33:45 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.20,471,1444719600"; d="scan'208";a="713839661" Received: from fmsmsx107.amr.corp.intel.com ([10.18.124.205]) by orsmga003.jf.intel.com with ESMTP; 23 Dec 2015 17:33:45 -0800 Received: from fmsmsx124.amr.corp.intel.com (10.18.125.39) by fmsmsx107.amr.corp.intel.com (10.18.124.205) with Microsoft SMTP Server (TLS) id 14.3.248.2; Wed, 23 Dec 2015 17:33:45 -0800 Received: from shsmsx102.ccr.corp.intel.com (10.239.4.154) by fmsmsx124.amr.corp.intel.com (10.18.125.39) with Microsoft SMTP Server (TLS) id 14.3.248.2; Wed, 23 Dec 2015 17:33:45 -0800 Received: from shsmsx101.ccr.corp.intel.com ([169.254.1.190]) by shsmsx102.ccr.corp.intel.com ([169.254.2.158]) with mapi id 14.03.0248.002; Thu, 24 Dec 2015 09:33:43 +0800 From: "Xie, Huawei" To: "Ananyev, Konstantin" , Stephen Hemminger Thread-Topic: [dpdk-dev] [PATCH v3 1/2] mbuf: provide rte_pktmbuf_alloc_bulk API Thread-Index: AdE96x9IC+XUeUndTwaFsXBlWnxwsw== Date: Thu, 24 Dec 2015 01:33:42 +0000 Message-ID: References: <1450055682-51953-1-git-send-email-huawei.xie@intel.com> <1450801074-29361-1-git-send-email-huawei.xie@intel.com> <1450801074-29361-2-git-send-email-huawei.xie@intel.com> <20151223103744.28551da7@xeon-e3> <2601191342CEEE43887BDE71AB97725836AD8261@irsmsx105.ger.corp.intel.com> Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [10.239.127.40] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Cc: "dev@dpdk.org" , "dprovan@bivio.net" Subject: Re: [dpdk-dev] [PATCH v3 1/2] mbuf: provide rte_pktmbuf_alloc_bulk API X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 24 Dec 2015 01:33:47 -0000 On 12/24/2015 2:49 AM, Ananyev, Konstantin wrote:=0A= >=0A= >> -----Original Message-----=0A= >> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Stephen Hemminger= =0A= >> Sent: Wednesday, December 23, 2015 6:38 PM=0A= >> To: Xie, Huawei=0A= >> Cc: dev@dpdk.org; dprovan@bivio.net=0A= >> Subject: Re: [dpdk-dev] [PATCH v3 1/2] mbuf: provide rte_pktmbuf_alloc_b= ulk API=0A= >>=0A= >> On Wed, 23 Dec 2015 00:17:53 +0800=0A= >> Huawei Xie wrote:=0A= >>=0A= >>> +=0A= >>> + rc =3D rte_mempool_get_bulk(pool, (void **)mbufs, count);=0A= >>> + if (unlikely(rc))=0A= >>> + return rc;=0A= >>> +=0A= >>> + switch (count % 4) {=0A= >>> + case 0: while (idx !=3D count) {=0A= >>> + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) =3D=3D 0);=0A= >>> + rte_mbuf_refcnt_set(mbufs[idx], 1);=0A= >>> + rte_pktmbuf_reset(mbufs[idx]);=0A= >>> + idx++;=0A= >>> + case 3:=0A= >>> + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) =3D=3D 0);=0A= >>> + rte_mbuf_refcnt_set(mbufs[idx], 1);=0A= >>> + rte_pktmbuf_reset(mbufs[idx]);=0A= >>> + idx++;=0A= >>> + case 2:=0A= >>> + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) =3D=3D 0);=0A= >>> + rte_mbuf_refcnt_set(mbufs[idx], 1);=0A= >>> + rte_pktmbuf_reset(mbufs[idx]);=0A= >>> + idx++;=0A= >>> + case 1:=0A= >>> + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) =3D=3D 0);=0A= >>> + rte_mbuf_refcnt_set(mbufs[idx], 1);=0A= >>> + rte_pktmbuf_reset(mbufs[idx]);=0A= >>> + idx++;=0A= >>> + }=0A= >>> + }=0A= >>> + return 0;=0A= >>> +}=0A= >> Since function will not work if count can not be 0 (otherwise rte_mempoo= l_get_bulk will fail),=0A= > As I understand, rte_mempool_get_bulk() will work correctly and return 0,= if count=3D=3D0.=0A= > That's why Huawei prefers while() {}, instead of do {} while() - to avoid= extra check for=0A= > (count !=3D 0) at the start. =0A= > Konstantin=0A= =0A= Yes.=0A= =0A= >=0A= >=0A= >> why not:=0A= >> 1. Document that assumption=0A= >> 2. Use that assumption to speed up code.=0A= >>=0A= >>=0A= >>=0A= >> switch(count % 4) {=0A= >> do {=0A= >> case 0:=0A= >> ...=0A= >> case 1:=0A= >> ...=0A= >> } while (idx !=3D count);=0A= >> }=0A= >>=0A= >> Also you really need to add a big block comment about this loop, to expl= ain=0A= >> what it does and why.=0A= =0A= Since we change duff's implementation a bit, and for people who don't=0A= know duff's device, we could add comment.=0A= Is comment like this enough?=0A= Use Duff's device to unroll the loop a bit to gain more performance=0A= Use while() rather than do {} while() as count could be zero.=0A= =0A= =0A=