Patch Detail
get:
Show a patch.
patch:
Update a patch.
put:
Update a patch.
GET /api/patches/81074/?format=api
{ "id": 81074, "url": "http://patches.dpdk.org/api/patches/81074/?format=api", "web_url": "http://patches.dpdk.org/project/dpdk/patch/20201016094431.96889-3-jia.guo@intel.com/", "project": { "id": 1, "url": "http://patches.dpdk.org/api/projects/1/?format=api", "name": "DPDK", "link_name": "dpdk", "list_id": "dev.dpdk.org", "list_email": "dev@dpdk.org", "web_url": "http://core.dpdk.org", "scm_url": "git://dpdk.org/dpdk", "webscm_url": "http://git.dpdk.org/dpdk", "list_archive_url": "https://inbox.dpdk.org/dev", "list_archive_url_format": "https://inbox.dpdk.org/dev/{}", "commit_url_format": "" }, "msgid": "<20201016094431.96889-3-jia.guo@intel.com>", "list_archive_url": "https://inbox.dpdk.org/dev/20201016094431.96889-3-jia.guo@intel.com", "date": "2020-10-16T09:44:28", "name": "[v5,2/5] net/i40e: fix vector rx burst for i40e", "commit_ref": null, "pull_url": null, "state": "accepted", "archived": true, "hash": "23b39d73552041ef36c061edb72319ff00faa3a1", "submitter": { "id": 507, "url": "http://patches.dpdk.org/api/people/507/?format=api", "name": "Guo, Jia", "email": "jia.guo@intel.com" }, "delegate": { "id": 1540, "url": "http://patches.dpdk.org/api/users/1540/?format=api", "username": "qzhan15", "first_name": "Qi", "last_name": "Zhang", "email": "qi.z.zhang@intel.com" }, "mbox": "http://patches.dpdk.org/project/dpdk/patch/20201016094431.96889-3-jia.guo@intel.com/mbox/", "series": [ { "id": 13053, "url": "http://patches.dpdk.org/api/series/13053/?format=api", "web_url": "http://patches.dpdk.org/project/dpdk/list/?series=13053", "date": "2020-10-16T09:44:26", "name": "fix vector rx burst for PMDs", "version": 5, "mbox": "http://patches.dpdk.org/series/13053/mbox/" } ], "comments": "http://patches.dpdk.org/api/patches/81074/comments/", "check": "success", "checks": "http://patches.dpdk.org/api/patches/81074/checks/", "tags": {}, "related": [], "headers": { "Return-Path": "<dev-bounces@dpdk.org>", "X-Original-To": "patchwork@inbox.dpdk.org", "Delivered-To": "patchwork@inbox.dpdk.org", "Received": [ "from dpdk.org (dpdk.org [92.243.14.124])\n\tby inbox.dpdk.org (Postfix) with ESMTP id BB024A04DB;\n\tFri, 16 Oct 2020 11:51:54 +0200 (CEST)", "from [92.243.14.124] (localhost [127.0.0.1])\n\tby dpdk.org (Postfix) with ESMTP id 307C31EC69;\n\tFri, 16 Oct 2020 11:51:18 +0200 (CEST)", "from mga14.intel.com (mga14.intel.com [192.55.52.115])\n by dpdk.org (Postfix) with ESMTP id DE44D1EC61\n for <dev@dpdk.org>; Fri, 16 Oct 2020 11:51:15 +0200 (CEST)", "from orsmga004.jf.intel.com ([10.7.209.38])\n by fmsmga103.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384;\n 16 Oct 2020 02:51:15 -0700", "from npg-dpdk-cvl-jeffguo-01.sh.intel.com ([10.67.111.128])\n by orsmga004.jf.intel.com with ESMTP; 16 Oct 2020 02:51:11 -0700" ], "IronPort-SDR": [ "\n tS43tSSyw4bCqDEjpMulJOWBI8/72luGlUfB3G0FvYtrZLAsl2IFv4h+cM4rHGZLFpua3r0vwg\n oNgHdDhu0plQ==", "\n UFGbDYJi5Np9V2fzKC5eCGQ2JHo16sXArbwIL75HDQROukln+fk/n5vzSa1/qRhYU/njviIIcb\n NgZAOlSlBQkw==" ], "X-IronPort-AV": [ "E=McAfee;i=\"6000,8403,9775\"; a=\"165806234\"", "E=Sophos;i=\"5.77,382,1596524400\"; d=\"scan'208\";a=\"165806234\"", "E=Sophos;i=\"5.77,382,1596524400\"; d=\"scan'208\";a=\"464629455\"" ], "X-Amp-Result": "SKIPPED(no attachment in message)", "X-Amp-File-Uploaded": "False", "X-ExtLoop1": "1", "From": "Jeff Guo <jia.guo@intel.com>", "To": "jingjing.wu@intel.com, qi.z.zhang@intel.com, beilei.xing@intel.com,\n haiyue.wang@intel.com, qiming.yang@intel.com", "Cc": "dev@dpdk.org, ferruh.yigit@intel.com, mb@smartsharesystems.com,\n stephen@networkplumber.org, barbette@kth.se, Feifei.wang2@arm.com,\n bruce.richardson@intel.com, jia.guo@intel.com, helin.zhang@intel.com", "Date": "Fri, 16 Oct 2020 17:44:28 +0800", "Message-Id": "<20201016094431.96889-3-jia.guo@intel.com>", "X-Mailer": "git-send-email 2.20.1", "In-Reply-To": "<20201016094431.96889-1-jia.guo@intel.com>", "References": "<20200827075452.1751-1-jia.guo@intel.com>\n <20201016094431.96889-1-jia.guo@intel.com>", "MIME-Version": "1.0", "Content-Type": "text/plain; charset=UTF-8", "Content-Transfer-Encoding": "8bit", "Subject": "[dpdk-dev] [PATCH v5 2/5] net/i40e: fix vector rx burst for i40e", "X-BeenThere": "dev@dpdk.org", "X-Mailman-Version": "2.1.15", "Precedence": "list", "List-Id": "DPDK patches and discussions <dev.dpdk.org>", "List-Unsubscribe": "<https://mails.dpdk.org/options/dev>,\n <mailto:dev-request@dpdk.org?subject=unsubscribe>", "List-Archive": "<http://mails.dpdk.org/archives/dev/>", "List-Post": "<mailto:dev@dpdk.org>", "List-Help": "<mailto:dev-request@dpdk.org?subject=help>", "List-Subscribe": "<https://mails.dpdk.org/listinfo/dev>,\n <mailto:dev-request@dpdk.org?subject=subscribe>", "Errors-To": "dev-bounces@dpdk.org", "Sender": "\"dev\" <dev-bounces@dpdk.org>" }, "content": "The limitation of burst size in vector rx was removed, since it should\nretrieve as much received packets as possible. And also the scattered\nreceive path should use a wrapper function to achieve the goal of\nburst maximizing.\n\nBugzilla ID: 516\nFixes: 5b463eda8d26 (\"net/i40e: make vector driver filenames consistent\")\nFixes: ae0eb310f253 (\"net/i40e: implement vector PMD for ARM\")\nFixes: c3def6a8724c (\"net/i40e: implement vector PMD for altivec\")\n\nSigned-off-by: Jeff Guo <jia.guo@intel.com>\nAcked-by: Morten Brørup <mb@smartsharesystems.com>\n---\n drivers/net/i40e/i40e_rxtx_vec_altivec.c | 59 +++++++++++++++++-------\n drivers/net/i40e/i40e_rxtx_vec_neon.c | 48 ++++++++++++++-----\n drivers/net/i40e/i40e_rxtx_vec_sse.c | 48 ++++++++++++++-----\n 3 files changed, 114 insertions(+), 41 deletions(-)", "diff": "diff --git a/drivers/net/i40e/i40e_rxtx_vec_altivec.c b/drivers/net/i40e/i40e_rxtx_vec_altivec.c\nindex 6862a017e1..d3238bfb6a 100644\n--- a/drivers/net/i40e/i40e_rxtx_vec_altivec.c\n+++ b/drivers/net/i40e/i40e_rxtx_vec_altivec.c\n@@ -188,11 +188,13 @@ desc_to_ptype_v(vector unsigned long descs[4], struct rte_mbuf **rx_pkts,\n \t\tptype_tbl[(*(vector unsigned char *)&ptype1)[8]];\n }\n \n- /* Notice:\n- * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n- */\n+/**\n+ * vPMD raw receive routine, only accept(nb_pkts >= RTE_I40E_DESCS_PER_LOOP)\n+ *\n+ * Notice:\n+ * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n+ * - floor align nb_pkts to a RTE_I40E_DESCS_PER_LOOP power-of-two\n+ */\n static inline uint16_t\n _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,\n \t\t uint16_t nb_pkts, uint8_t *split_packet)\n@@ -214,9 +216,6 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,\n \t\t};\n \tvector unsigned long dd_check, eop_check;\n \n-\t/* nb_pkts shall be less equal than RTE_I40E_MAX_RX_BURST */\n-\tnb_pkts = RTE_MIN(nb_pkts, RTE_I40E_MAX_RX_BURST);\n-\n \t/* nb_pkts has to be floor-aligned to RTE_I40E_DESCS_PER_LOOP */\n \tnb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_I40E_DESCS_PER_LOOP);\n \n@@ -459,15 +458,15 @@ i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n \treturn _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);\n }\n \n- /* vPMD receive routine that reassembles scattered packets\n- * Notice:\n- * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n- */\n-uint16_t\n-i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n-\t\t\t uint16_t nb_pkts)\n+/**\n+ * vPMD receive routine that reassembles single burst of 32 scattered packets\n+ *\n+ * Notice:\n+ * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n+ */\n+static uint16_t\n+i40e_recv_scattered_burst_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n {\n \tstruct i40e_rx_queue *rxq = rx_queue;\n \tuint8_t split_flags[RTE_I40E_VPMD_RX_BURST] = {0};\n@@ -500,6 +499,32 @@ i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n \t\t&split_flags[i]);\n }\n \n+/**\n+ * vPMD receive routine that reassembles scattered packets.\n+ */\n+uint16_t\n+i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n+{\n+\tuint16_t retval = 0;\n+\n+\twhile (nb_pkts > RTE_I40E_VPMD_RX_BURST) {\n+\t\tuint16_t burst;\n+\n+\t\tburst = i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t RTE_I40E_VPMD_RX_BURST);\n+\t\tretval += burst;\n+\t\tnb_pkts -= burst;\n+\t\tif (burst < RTE_I40E_VPMD_RX_BURST)\n+\t\t\treturn retval;\n+\t}\n+\n+\treturn retval + i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t nb_pkts);\n+}\n+\n static inline void\n vtx1(volatile struct i40e_tx_desc *txdp,\n \tstruct rte_mbuf *pkt, uint64_t flags)\ndiff --git a/drivers/net/i40e/i40e_rxtx_vec_neon.c b/drivers/net/i40e/i40e_rxtx_vec_neon.c\nindex 543ecadb07..f094de69ae 100644\n--- a/drivers/net/i40e/i40e_rxtx_vec_neon.c\n+++ b/drivers/net/i40e/i40e_rxtx_vec_neon.c\n@@ -187,11 +187,12 @@ desc_to_ptype_v(uint64x2_t descs[4], struct rte_mbuf **__rte_restrict rx_pkts,\n \n }\n \n- /*\n+/**\n+ * vPMD raw receive routine, only accept(nb_pkts >= RTE_I40E_DESCS_PER_LOOP)\n+ *\n * Notice:\n * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n+ * - floor align nb_pkts to a RTE_I40E_DESCS_PER_LOOP power-of-two\n */\n static inline uint16_t\n _recv_raw_pkts_vec(struct i40e_rx_queue *__rte_restrict rxq,\n@@ -230,9 +231,6 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *__rte_restrict rxq,\n \t\t0, 0, 0 /* ignore non-length fields */\n \t\t};\n \n-\t/* nb_pkts shall be less equal than RTE_I40E_MAX_RX_BURST */\n-\tnb_pkts = RTE_MIN(nb_pkts, RTE_I40E_MAX_RX_BURST);\n-\n \t/* nb_pkts has to be floor-aligned to RTE_I40E_DESCS_PER_LOOP */\n \tnb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_I40E_DESCS_PER_LOOP);\n \n@@ -439,15 +437,15 @@ i40e_recv_pkts_vec(void *__rte_restrict rx_queue,\n \treturn _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);\n }\n \n- /* vPMD receive routine that reassembles scattered packets\n+/**\n+ * vPMD receive routine that reassembles single burst of 32 scattered packets\n+ *\n * Notice:\n * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n */\n-uint16_t\n-i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n-\t\t\t uint16_t nb_pkts)\n+static uint16_t\n+i40e_recv_scattered_burst_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n {\n \n \tstruct i40e_rx_queue *rxq = rx_queue;\n@@ -482,6 +480,32 @@ i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n \t\t&split_flags[i]);\n }\n \n+/**\n+ * vPMD receive routine that reassembles scattered packets.\n+ */\n+uint16_t\n+i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n+{\n+\tuint16_t retval = 0;\n+\n+\twhile (nb_pkts > RTE_I40E_VPMD_RX_BURST) {\n+\t\tuint16_t burst;\n+\n+\t\tburst = i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t RTE_I40E_VPMD_RX_BURST);\n+\t\tretval += burst;\n+\t\tnb_pkts -= burst;\n+\t\tif (burst < RTE_I40E_VPMD_RX_BURST)\n+\t\t\treturn retval;\n+\t}\n+\n+\treturn retval + i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t nb_pkts);\n+}\n+\n static inline void\n vtx1(volatile struct i40e_tx_desc *txdp,\n \t\tstruct rte_mbuf *pkt, uint64_t flags)\ndiff --git a/drivers/net/i40e/i40e_rxtx_vec_sse.c b/drivers/net/i40e/i40e_rxtx_vec_sse.c\nindex 240ce478ab..4b2b6a28fc 100644\n--- a/drivers/net/i40e/i40e_rxtx_vec_sse.c\n+++ b/drivers/net/i40e/i40e_rxtx_vec_sse.c\n@@ -342,11 +342,12 @@ desc_to_ptype_v(__m128i descs[4], struct rte_mbuf **rx_pkts,\n \trx_pkts[3]->packet_type = ptype_tbl[_mm_extract_epi8(ptype1, 8)];\n }\n \n- /*\n+/**\n+ * vPMD raw receive routine, only accept(nb_pkts >= RTE_I40E_DESCS_PER_LOOP)\n+ *\n * Notice:\n * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n+ * - floor align nb_pkts to a RTE_I40E_DESCS_PER_LOOP power-of-two\n */\n static inline uint16_t\n _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,\n@@ -378,9 +379,6 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_pkts,\n \t\t\toffsetof(struct rte_mbuf, rx_descriptor_fields1) + 8);\n \t__m128i dd_check, eop_check;\n \n-\t/* nb_pkts shall be less equal than RTE_I40E_MAX_RX_BURST */\n-\tnb_pkts = RTE_MIN(nb_pkts, RTE_I40E_MAX_RX_BURST);\n-\n \t/* nb_pkts has to be floor-aligned to RTE_I40E_DESCS_PER_LOOP */\n \tnb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_I40E_DESCS_PER_LOOP);\n \n@@ -605,15 +603,15 @@ i40e_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n \treturn _recv_raw_pkts_vec(rx_queue, rx_pkts, nb_pkts, NULL);\n }\n \n- /* vPMD receive routine that reassembles scattered packets\n+/**\n+ * vPMD receive routine that reassembles single burst of 32 scattered packets\n+ *\n * Notice:\n * - nb_pkts < RTE_I40E_DESCS_PER_LOOP, just return no packet\n- * - nb_pkts > RTE_I40E_VPMD_RX_BURST, only scan RTE_I40E_VPMD_RX_BURST\n- * numbers of DD bits\n */\n-uint16_t\n-i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n-\t\t\t uint16_t nb_pkts)\n+static uint16_t\n+i40e_recv_scattered_burst_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n {\n \n \tstruct i40e_rx_queue *rxq = rx_queue;\n@@ -648,6 +646,32 @@ i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n \t\t&split_flags[i]);\n }\n \n+/**\n+ * vPMD receive routine that reassembles scattered packets.\n+ */\n+uint16_t\n+i40e_recv_scattered_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,\n+\t\t\t uint16_t nb_pkts)\n+{\n+\tuint16_t retval = 0;\n+\n+\twhile (nb_pkts > RTE_I40E_VPMD_RX_BURST) {\n+\t\tuint16_t burst;\n+\n+\t\tburst = i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t RTE_I40E_VPMD_RX_BURST);\n+\t\tretval += burst;\n+\t\tnb_pkts -= burst;\n+\t\tif (burst < RTE_I40E_VPMD_RX_BURST)\n+\t\t\treturn retval;\n+\t}\n+\n+\treturn retval + i40e_recv_scattered_burst_vec(rx_queue,\n+\t\t\t\t\t\t rx_pkts + retval,\n+\t\t\t\t\t\t nb_pkts);\n+}\n+\n static inline void\n vtx1(volatile struct i40e_tx_desc *txdp,\n \t\tstruct rte_mbuf *pkt, uint64_t flags)\n", "prefixes": [ "v5", "2/5" ] }