Patch Detail
get:
Show a patch.
patch:
Update a patch.
put:
Update a patch.
GET /api/patches/17250/?format=api
{ "id": 17250, "url": "https://patches.dpdk.org/api/patches/17250/?format=api", "web_url": "https://patches.dpdk.org/project/dpdk/patch/fb9b87ee0cdff03fb949fb8e0ea3ce52cae6bd86.1479995764.git.nelio.laranjeiro@6wind.com/", "project": { "id": 1, "url": "https://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": "<fb9b87ee0cdff03fb949fb8e0ea3ce52cae6bd86.1479995764.git.nelio.laranjeiro@6wind.com>", "list_archive_url": "https://inbox.dpdk.org/dev/fb9b87ee0cdff03fb949fb8e0ea3ce52cae6bd86.1479995764.git.nelio.laranjeiro@6wind.com", "date": "2016-11-24T16:03:32", "name": "[dpdk-dev,3/7] net/mlx5: use vector types to speed up processing", "commit_ref": null, "pull_url": null, "state": "accepted", "archived": true, "hash": "3d7b1c8ff409026385268d83e98cdce3a22b8e0b", "submitter": { "id": 243, "url": "https://patches.dpdk.org/api/people/243/?format=api", "name": "Nélio Laranjeiro", "email": "nelio.laranjeiro@6wind.com" }, "delegate": { "id": 319, "url": "https://patches.dpdk.org/api/users/319/?format=api", "username": "fyigit", "first_name": "Ferruh", "last_name": "Yigit", "email": "ferruh.yigit@amd.com" }, "mbox": "https://patches.dpdk.org/project/dpdk/patch/fb9b87ee0cdff03fb949fb8e0ea3ce52cae6bd86.1479995764.git.nelio.laranjeiro@6wind.com/mbox/", "series": [], "comments": "https://patches.dpdk.org/api/patches/17250/comments/", "check": "success", "checks": "https://patches.dpdk.org/api/patches/17250/checks/", "tags": {}, "related": [], "headers": { "Return-Path": "<dev-bounces@dpdk.org>", "X-Original-To": "patchwork@dpdk.org", "Delivered-To": "patchwork@dpdk.org", "Received": [ "from [92.243.14.124] (localhost [IPv6:::1])\n\tby dpdk.org (Postfix) with ESMTP id E284C5681;\n\tThu, 24 Nov 2016 17:04:17 +0100 (CET)", "from mail-wm0-f51.google.com (mail-wm0-f51.google.com\n\t[74.125.82.51]) by dpdk.org (Postfix) with ESMTP id 382012BEF\n\tfor <dev@dpdk.org>; Thu, 24 Nov 2016 17:04:03 +0100 (CET)", "by mail-wm0-f51.google.com with SMTP id t79so66078534wmt.0\n\tfor <dev@dpdk.org>; Thu, 24 Nov 2016 08:04:03 -0800 (PST)", "from ping.vm.6wind.com (guy78-3-82-239-227-177.fbx.proxad.net.\n\t[82.239.227.177]) by smtp.gmail.com with ESMTPSA id\n\tvr9sm42495142wjc.35.2016.11.24.08.04.01\n\t(version=TLS1_2 cipher=ECDHE-RSA-AES128-SHA bits=128/128);\n\tThu, 24 Nov 2016 08:04:02 -0800 (PST)" ], "DKIM-Signature": "v=1; a=rsa-sha256; c=relaxed/relaxed;\n\td=6wind-com.20150623.gappssmtp.com; s=20150623;\n\th=from:to:cc:subject:date:message-id:in-reply-to:references\n\t:in-reply-to:references;\n\tbh=/X1WR7n1vNmsIthnGA/BKfPnQjFmapCVYYMT8MzEpVM=;\n\tb=HD10RUoufSOJYKsq1vUZEF47XOO1VFEKyiOkdvPeFQQZP/hoMyOjL/KcjfRvR+KbGL\n\tZyvgsbDZ3UTlkD08thGhDkCdD/zNLS4XRUJDV0rqPvCW7IJrr7yBlL2xxaHGvq+eFMcY\n\thvdM7VaNCjkgXQKYcf+CbU4vePtlEpOl0uEjtlM4wVm2W/O4e0WJn68ruris8UwauhB3\n\t9aI1XTdwcAXIXDQw/zG+rmyQgm8dkon88ew2AE1/2jSV/yYXdInETo6pJjYrnNrNVB/S\n\tO5BKTXG1e3bgqgMF3GzgTd0SuWiE0627xTyKy0FBZadHiqsRD69z1P7zly2K7DmyLzPL\n\t30nQ==", "X-Google-DKIM-Signature": "v=1; a=rsa-sha256; c=relaxed/relaxed;\n\td=1e100.net; s=20130820;\n\th=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to\n\t:references:in-reply-to:references;\n\tbh=/X1WR7n1vNmsIthnGA/BKfPnQjFmapCVYYMT8MzEpVM=;\n\tb=ZgSG5HyOffEDJW38KnixVL8wAx6Q6quO6DTxxmFn+OrxggYRZB6uPNz2Dk+5lmJi+9\n\tiEZZMWxzXRVgSzjSHvjmyMkka/32fLGKjyEolLapYjpkCdXI8c1i/gPv5YDb7qjLNciM\n\tSxyEb2PCaB/MZ4Ik3hvqfendpT6tn6SZ07YtsNkqJEeVCW+QRp+KBqfv1aTQbtlPmhnr\n\tySldXBkbwDQV42MYFNF+n7PkkPBV29EtvUFYepKcfCt+D1wljEYxDNTCuYIemaw5RKj1\n\tZADkVeDjZOupSGZiMwYb08FOmcPldpUOy7OlevHQcWulcRMJgyfNU5cSRO6SJaM0z8HK\n\tqVeA==", "X-Gm-Message-State": "AKaTC02i+qR2xy2QzWKICclTfL4F0TUzb3HUZsev2PonEGR28MbfsvHD9dC74t2i6v23ov20", "X-Received": "by 10.28.22.193 with SMTP id 184mr3087538wmw.100.1480003442826; \n\tThu, 24 Nov 2016 08:04:02 -0800 (PST)", "From": "Nelio Laranjeiro <nelio.laranjeiro@6wind.com>", "To": "dev@dpdk.org", "Cc": "Thomas Monjalon <thomas.monjalon@6wind.com>,\n\tAdrien Mazarguil <adrien.mazarguil@6wind.com>", "Date": "Thu, 24 Nov 2016 17:03:32 +0100", "Message-Id": "<fb9b87ee0cdff03fb949fb8e0ea3ce52cae6bd86.1479995764.git.nelio.laranjeiro@6wind.com>", "X-Mailer": "git-send-email 2.1.4", "In-Reply-To": [ "<cover.1479995764.git.nelio.laranjeiro@6wind.com>", "<cover.1479995764.git.nelio.laranjeiro@6wind.com>" ], "References": [ "<cover.1479995764.git.nelio.laranjeiro@6wind.com>", "<cover.1479995764.git.nelio.laranjeiro@6wind.com>" ], "Subject": "[dpdk-dev] [PATCH 3/7] net/mlx5: use vector types to speed up\n\tprocessing", "X-BeenThere": "dev@dpdk.org", "X-Mailman-Version": "2.1.15", "Precedence": "list", "List-Id": "patches and discussions about DPDK <dev.dpdk.org>", "List-Unsubscribe": "<http://dpdk.org/ml/options/dev>,\n\t<mailto:dev-request@dpdk.org?subject=unsubscribe>", "List-Archive": "<http://dpdk.org/ml/archives/dev/>", "List-Post": "<mailto:dev@dpdk.org>", "List-Help": "<mailto:dev-request@dpdk.org?subject=help>", "List-Subscribe": "<http://dpdk.org/ml/listinfo/dev>,\n\t<mailto:dev-request@dpdk.org?subject=subscribe>", "Errors-To": "dev-bounces@dpdk.org", "Sender": "\"dev\" <dev-bounces@dpdk.org>" }, "content": "Let compiler automatically use the vector capabilities of the target\nmachine to optimize instructions.\n\nSigned-off-by: Nelio Laranjeiro <nelio.laranjeiro@6wind.com>\nAcked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>\n---\n drivers/net/mlx5/mlx5_prm.h | 7 +++++\n drivers/net/mlx5/mlx5_rxtx.c | 74 +++++++++++++++++++++++---------------------\n 2 files changed, 46 insertions(+), 35 deletions(-)", "diff": "diff --git a/drivers/net/mlx5/mlx5_prm.h b/drivers/net/mlx5/mlx5_prm.h\nindex 3dd4cbe..9cd9fdf 100644\n--- a/drivers/net/mlx5/mlx5_prm.h\n+++ b/drivers/net/mlx5/mlx5_prm.h\n@@ -44,6 +44,7 @@\n #pragma GCC diagnostic error \"-Wpedantic\"\n #endif\n \n+#include <rte_vect.h>\n #include \"mlx5_autoconf.h\"\n \n /* Get CQE owner bit. */\n@@ -134,6 +135,12 @@ struct mlx5_wqe {\n \tstruct mlx5_wqe_eth_seg_small eseg;\n };\n \n+/* Vectorize WQE header. */\n+struct mlx5_wqe_v {\n+\trte_v128u32_t ctrl;\n+\trte_v128u32_t eseg;\n+};\n+\n /* WQE. */\n struct mlx5_wqe64 {\n \tstruct mlx5_wqe hdr;\ndiff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c\nindex ada8e74..e161cd9 100644\n--- a/drivers/net/mlx5/mlx5_rxtx.c\n+++ b/drivers/net/mlx5/mlx5_rxtx.c\n@@ -371,7 +371,7 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \tunsigned int j = 0;\n \tunsigned int max;\n \tunsigned int comp;\n-\tvolatile struct mlx5_wqe *wqe = NULL;\n+\tvolatile struct mlx5_wqe_v *wqe = NULL;\n \tunsigned int segs_n = 0;\n \tstruct rte_mbuf *buf = NULL;\n \tuint8_t *raw;\n@@ -388,12 +388,13 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \tif (max > elts_n)\n \t\tmax -= elts_n;\n \tdo {\n-\t\tvolatile struct mlx5_wqe_data_seg *dseg = NULL;\n+\t\tvolatile rte_v128u32_t *dseg = NULL;\n \t\tuint32_t length;\n \t\tunsigned int ds = 0;\n \t\tuintptr_t addr;\n \t\tuint16_t pkt_inline_sz = MLX5_WQE_DWORD_SIZE;\n \t\tuint8_t ehdr[2];\n+\t\tuint8_t cs_flags = 0;\n #ifdef MLX5_PMD_SOFT_COUNTERS\n \t\tuint32_t total_length = 0;\n #endif\n@@ -412,7 +413,7 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\t--segs_n;\n \t\tif (!segs_n)\n \t\t\t--pkts_n;\n-\t\twqe = (volatile struct mlx5_wqe *)\n+\t\twqe = (volatile struct mlx5_wqe_v *)\n \t\t\ttx_mlx5_wqe(txq, txq->wqe_ci);\n \t\trte_prefetch0(tx_mlx5_wqe(txq, txq->wqe_ci + 1));\n \t\tif (pkts_n > 1)\n@@ -438,11 +439,7 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\t/* Should we enable HW CKSUM offload */\n \t\tif (buf->ol_flags &\n \t\t (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) {\n-\t\t\twqe->eseg.cs_flags =\n-\t\t\t\tMLX5_ETH_WQE_L3_CSUM |\n-\t\t\t\tMLX5_ETH_WQE_L4_CSUM;\n-\t\t} else {\n-\t\t\twqe->eseg.cs_flags = 0;\n+\t\t\tcs_flags = MLX5_ETH_WQE_L3_CSUM | MLX5_ETH_WQE_L4_CSUM;\n \t\t}\n \t\traw = ((uint8_t *)(uintptr_t)wqe) + 2 * MLX5_WQE_DWORD_SIZE;\n \t\t/*\n@@ -498,12 +495,11 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\t\t */\n \t\t\tds = 2 + MLX5_WQE_DS(pkt_inline_sz - 2);\n \t\t\tif (length > 0) {\n-\t\t\t\tdseg = (volatile struct mlx5_wqe_data_seg *)\n+\t\t\t\tdseg = (volatile rte_v128u32_t *)\n \t\t\t\t\t((uintptr_t)wqe +\n \t\t\t\t\t (ds * MLX5_WQE_DWORD_SIZE));\n \t\t\t\tif ((uintptr_t)dseg >= end)\n-\t\t\t\t\tdseg = (volatile struct\n-\t\t\t\t\t\tmlx5_wqe_data_seg *)\n+\t\t\t\t\tdseg = (volatile rte_v128u32_t *)\n \t\t\t\t\t txq->wqes;\n \t\t\t\tgoto use_dseg;\n \t\t\t} else if (!segs_n) {\n@@ -516,16 +512,17 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\t\t * No inline has been done in the packet, only the\n \t\t\t * Ethernet Header as been stored.\n \t\t\t */\n-\t\t\twqe->eseg.inline_hdr_sz = htons(MLX5_WQE_DWORD_SIZE);\n-\t\t\tdseg = (volatile struct mlx5_wqe_data_seg *)\n+\t\t\tdseg = (volatile rte_v128u32_t *)\n \t\t\t\t((uintptr_t)wqe + (3 * MLX5_WQE_DWORD_SIZE));\n \t\t\tds = 3;\n use_dseg:\n \t\t\t/* Add the remaining packet as a simple ds. */\n-\t\t\t*dseg = (volatile struct mlx5_wqe_data_seg) {\n-\t\t\t\t.addr = htonll(addr),\n-\t\t\t\t.byte_count = htonl(length),\n-\t\t\t\t.lkey = txq_mp2mr(txq, txq_mb2mp(buf)),\n+\t\t\taddr = htonll(addr);\n+\t\t\t*dseg = (rte_v128u32_t){\n+\t\t\t\thtonl(length),\n+\t\t\t\ttxq_mp2mr(txq, txq_mb2mp(buf)),\n+\t\t\t\taddr,\n+\t\t\t\taddr >> 32,\n \t\t\t};\n \t\t\t++ds;\n \t\t\tif (!segs_n)\n@@ -545,7 +542,7 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\t\tunsigned int n = (txq->wqe_ci + ((ds + 3) / 4)) &\n \t\t\t\t((1 << txq->wqe_n) - 1);\n \n-\t\t\tdseg = (volatile struct mlx5_wqe_data_seg *)\n+\t\t\tdseg = (volatile rte_v128u32_t *)\n \t\t\t tx_mlx5_wqe(txq, n);\n \t\t\trte_prefetch0(tx_mlx5_wqe(txq, n + 1));\n \t\t} else {\n@@ -559,10 +556,12 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t\ttotal_length += length;\n #endif\n \t\t/* Store segment information. */\n-\t\t*dseg = (volatile struct mlx5_wqe_data_seg) {\n-\t\t\t.addr = htonll(rte_pktmbuf_mtod(buf, uintptr_t)),\n-\t\t\t.byte_count = htonl(length),\n-\t\t\t.lkey = txq_mp2mr(txq, txq_mb2mp(buf)),\n+\t\taddr = htonll(rte_pktmbuf_mtod(buf, uintptr_t));\n+\t\t*dseg = (rte_v128u32_t){\n+\t\t\thtonl(length),\n+\t\t\ttxq_mp2mr(txq, txq_mb2mp(buf)),\n+\t\t\taddr,\n+\t\t\taddr >> 32,\n \t\t};\n \t\t(*txq->elts)[elts_head] = buf;\n \t\telts_head = (elts_head + 1) & (elts_n - 1);\n@@ -575,17 +574,19 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n next_pkt:\n \t\t++i;\n \t\t/* Initialize known and common part of the WQE structure. */\n-\t\twqe->ctrl[0] = htonl((txq->wqe_ci << 8) | MLX5_OPCODE_SEND);\n-\t\twqe->ctrl[1] = htonl(txq->qp_num_8s | ds);\n-\t\twqe->ctrl[2] = 0;\n-\t\twqe->ctrl[3] = 0;\n-\t\twqe->eseg.rsvd0 = 0;\n-\t\twqe->eseg.rsvd1 = 0;\n-\t\twqe->eseg.mss = 0;\n-\t\twqe->eseg.rsvd2 = 0;\n-\t\twqe->eseg.inline_hdr_sz = htons(pkt_inline_sz);\n-\t\twqe->eseg.inline_hdr[0] = ehdr[0];\n-\t\twqe->eseg.inline_hdr[1] = ehdr[1];\n+\t\twqe->ctrl = (rte_v128u32_t){\n+\t\t\thtonl((txq->wqe_ci << 8) | MLX5_OPCODE_SEND),\n+\t\t\thtonl(txq->qp_num_8s | ds),\n+\t\t\t0,\n+\t\t\t0,\n+\t\t};\n+\t\twqe->eseg = (rte_v128u32_t){\n+\t\t\t0,\n+\t\t\tcs_flags,\n+\t\t\t0,\n+\t\t\t(ehdr[1] << 24) | (ehdr[0] << 16) |\n+\t\t\thtons(pkt_inline_sz),\n+\t\t};\n \t\ttxq->wqe_ci += (ds + 3) / 4;\n #ifdef MLX5_PMD_SOFT_COUNTERS\n \t\t/* Increment sent bytes counter. */\n@@ -598,10 +599,13 @@ mlx5_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n)\n \t/* Check whether completion threshold has been reached. */\n \tcomp = txq->elts_comp + i + j;\n \tif (comp >= MLX5_TX_COMP_THRESH) {\n+\t\tvolatile struct mlx5_wqe_ctrl *w =\n+\t\t\t(volatile struct mlx5_wqe_ctrl *)wqe;\n+\n \t\t/* Request completion on last WQE. */\n-\t\twqe->ctrl[2] = htonl(8);\n+\t\tw->ctrl2 = htonl(8);\n \t\t/* Save elts_head in unused \"immediate\" field of WQE. */\n-\t\twqe->ctrl[3] = elts_head;\n+\t\tw->ctrl3 = elts_head;\n \t\ttxq->elts_comp = 0;\n \t} else {\n \t\ttxq->elts_comp = comp;\n", "prefixes": [ "dpdk-dev", "3/7" ] }