Patch Detail
get:
Show a patch.
patch:
Update a patch.
put:
Update a patch.
GET /api/patches/86252/?format=api
{ "id": 86252, "url": "https://patches.dpdk.org/api/patches/86252/?format=api", "web_url": "https://patches.dpdk.org/project/dpdk/patch/20210110111023.9525-5-ophirmu@nvidia.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": "<20210110111023.9525-5-ophirmu@nvidia.com>", "list_archive_url": "https://inbox.dpdk.org/dev/20210110111023.9525-5-ophirmu@nvidia.com", "date": "2021-01-10T11:10:21", "name": "[v3,4/6] app/regex: support multi cores", "commit_ref": null, "pull_url": null, "state": "accepted", "archived": true, "hash": "237c7b3dfef5b7714bb2f785cf08aa9a24726417", "submitter": { "id": 1908, "url": "https://patches.dpdk.org/api/people/1908/?format=api", "name": "Ophir Munk", "email": "ophirmu@nvidia.com" }, "delegate": { "id": 1, "url": "https://patches.dpdk.org/api/users/1/?format=api", "username": "tmonjalo", "first_name": "Thomas", "last_name": "Monjalon", "email": "thomas@monjalon.net" }, "mbox": "https://patches.dpdk.org/project/dpdk/patch/20210110111023.9525-5-ophirmu@nvidia.com/mbox/", "series": [ { "id": 14620, "url": "https://patches.dpdk.org/api/series/14620/?format=api", "web_url": "https://patches.dpdk.org/project/dpdk/list/?series=14620", "date": "2021-01-10T11:10:19", "name": "regex multi Q with multi cores support", "version": 3, "mbox": "https://patches.dpdk.org/series/14620/mbox/" } ], "comments": "https://patches.dpdk.org/api/patches/86252/comments/", "check": "warning", "checks": "https://patches.dpdk.org/api/patches/86252/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 mails.dpdk.org (xvm-189-124.dc0.ghst.net [217.70.189.124])\n\tby inbox.dpdk.org (Postfix) with ESMTP id A9FACA09EF;\n\tSun, 10 Jan 2021 12:11:05 +0100 (CET)", "from [217.70.189.124] (localhost [127.0.0.1])\n\tby mails.dpdk.org (Postfix) with ESMTP id 8D503140F91;\n\tSun, 10 Jan 2021 12:10:33 +0100 (CET)", "from mellanox.co.il (mail-il-dmz.mellanox.com [193.47.165.129])\n by mails.dpdk.org (Postfix) with ESMTP id ACF7D140F69\n for <dev@dpdk.org>; Sun, 10 Jan 2021 12:10:30 +0100 (CET)", "from Internal Mail-Server by MTLPINE1 (envelope-from\n ophirmu@nvidia.com) with SMTP; 10 Jan 2021 13:10:25 +0200", "from nvidia.com (pegasus05.mtr.labs.mlnx [10.210.16.100])\n by labmailer.mlnx (8.13.8/8.13.8) with ESMTP id 10ABAOOn032562;\n Sun, 10 Jan 2021 13:10:25 +0200" ], "From": "Ophir Munk <ophirmu@nvidia.com>", "To": "dev@dpdk.org", "Cc": "Ori Kam <orika@nvidia.com>, Thomas Monjalon <thomas@monjalon.net>,\n Ophir Munk <ophirmu@mellanox.com>", "Date": "Sun, 10 Jan 2021 11:10:21 +0000", "Message-Id": "<20210110111023.9525-5-ophirmu@nvidia.com>", "X-Mailer": "git-send-email 2.8.4", "In-Reply-To": "<20210110111023.9525-1-ophirmu@nvidia.com>", "References": "<20201220104148.13961-2-ophirmu@nvidia.com>\n <20210110111023.9525-1-ophirmu@nvidia.com>", "Subject": "[dpdk-dev] [PATCH v3 4/6] app/regex: support multi cores", "X-BeenThere": "dev@dpdk.org", "X-Mailman-Version": "2.1.29", "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": "Up to this commit the regex application was running with multiple QPs on\na single core. This commit adds the option to specify a number of cores\non which multiple QPs will run.\nA new parameter 'nb_lcores' was added to configure the number of cores:\n--nb_lcores <num of cores>.\nIf not configured the number of cores is set to 1 by default. On\napplication startup a few initial steps occur by the main core: the\nnumber of QPs and cores are parsed. The QPs are distributed as evenly\nas possible on the cores. The regex device and all QPs are initialized.\nThe data file is read and saved in a buffer. Then for each core the\napplication calls rte_eal_remote_launch() with the worker routine\n(run_regex) as its parameter.\n\nSigned-off-by: Ophir Munk <ophirmu@nvidia.com>\nAcked-by: Ori Kam <orika@nvidia.com>\n---\n app/test-regex/main.c | 155 ++++++++++++++++++++++++++++++++++++-----\n doc/guides/tools/testregex.rst | 30 ++++++--\n 2 files changed, 164 insertions(+), 21 deletions(-)", "diff": "diff --git a/app/test-regex/main.c b/app/test-regex/main.c\nindex 9bafd02..2948d3e 100644\n--- a/app/test-regex/main.c\n+++ b/app/test-regex/main.c\n@@ -34,6 +34,7 @@ enum app_args {\n \tARG_PERF_MODE,\n \tARG_NUM_OF_ITERATIONS,\n \tARG_NUM_OF_QPS,\n+\tARG_NUM_OF_LCORES,\n };\n \n struct job_ctx {\n@@ -49,6 +50,26 @@ struct qp_params {\n \tchar *buf;\n };\n \n+struct qps_per_lcore {\n+\tunsigned int lcore_id;\n+\tint socket;\n+\tuint16_t qp_id_base;\n+\tuint16_t nb_qps;\n+};\n+\n+struct regex_conf {\n+\tuint32_t nb_jobs;\n+\tbool perf_mode;\n+\tuint32_t nb_iterations;\n+\tchar *data_file;\n+\tuint8_t nb_max_matches;\n+\tuint32_t nb_qps;\n+\tuint16_t qp_id_base;\n+\tchar *data_buf;\n+\tlong data_len;\n+\tlong job_len;\n+};\n+\n static void\n usage(const char *prog_name)\n {\n@@ -58,14 +79,15 @@ usage(const char *prog_name)\n \t\t\" --nb_jobs: number of jobs to use\\n\"\n \t\t\" --perf N: only outputs the performance data\\n\"\n \t\t\" --nb_iter N: number of iteration to run\\n\"\n-\t\t\" --nb_qps N: number of queues to use\\n\",\n+\t\t\" --nb_qps N: number of queues to use\\n\"\n+\t\t\" --nb_lcores N: number of lcores to use\\n\",\n \t\tprog_name);\n }\n \n static void\n args_parse(int argc, char **argv, char *rules_file, char *data_file,\n \t uint32_t *nb_jobs, bool *perf_mode, uint32_t *nb_iterations,\n-\t uint32_t *nb_qps)\n+\t uint32_t *nb_qps, uint32_t *nb_lcores)\n {\n \tchar **argvopt;\n \tint opt;\n@@ -85,6 +107,8 @@ args_parse(int argc, char **argv, char *rules_file, char *data_file,\n \t\t{ \"nb_iter\", 1, 0, ARG_NUM_OF_ITERATIONS},\n \t\t/* Number of QPs. */\n \t\t{ \"nb_qps\", 1, 0, ARG_NUM_OF_QPS},\n+\t\t/* Number of lcores. */\n+\t\t{ \"nb_lcores\", 1, 0, ARG_NUM_OF_LCORES},\n \t\t/* End of options */\n \t\t{ 0, 0, 0, 0 }\n \t};\n@@ -121,6 +145,9 @@ args_parse(int argc, char **argv, char *rules_file, char *data_file,\n \t\tcase ARG_NUM_OF_QPS:\n \t\t\t*nb_qps = atoi(optarg);\n \t\t\tbreak;\n+\t\tcase ARG_NUM_OF_LCORES:\n+\t\t\t*nb_lcores = atoi(optarg);\n+\t\t\tbreak;\n \t\tcase ARG_HELP:\n \t\t\tusage(\"RegEx test app\");\n \t\t\tbreak;\n@@ -274,11 +301,18 @@ extbuf_free_cb(void *addr __rte_unused, void *fcb_opaque __rte_unused)\n }\n \n static int\n-run_regex(uint32_t nb_jobs,\n-\t bool perf_mode, uint32_t nb_iterations,\n-\t uint8_t nb_max_matches, uint32_t nb_qps,\n-\t char *data_buf, long data_len, long job_len)\n+run_regex(void *args)\n {\n+\tstruct regex_conf *rgxc = args;\n+\tuint32_t nb_jobs = rgxc->nb_jobs;\n+\tuint32_t nb_iterations = rgxc->nb_iterations;\n+\tuint8_t nb_max_matches = rgxc->nb_max_matches;\n+\tuint32_t nb_qps = rgxc->nb_qps;\n+\tuint16_t qp_id_base = rgxc->qp_id_base;\n+\tchar *data_buf = rgxc->data_buf;\n+\tlong data_len = rgxc->data_len;\n+\tlong job_len = rgxc->job_len;\n+\n \tchar *buf = NULL;\n \tuint32_t actual_jobs = 0;\n \tuint32_t i;\n@@ -298,9 +332,13 @@ run_regex(uint32_t nb_jobs,\n \tstruct qp_params *qps = NULL;\n \tbool update;\n \tuint16_t qps_used = 0;\n+\tchar mbuf_pool[16];\n \n \tshinfo.free_cb = extbuf_free_cb;\n-\tmbuf_mp = rte_pktmbuf_pool_create(\"mbuf_pool\", nb_jobs * nb_qps, 0,\n+\tsnprintf(mbuf_pool,\n+\t\t sizeof(mbuf_pool),\n+\t\t \"mbuf_pool_%2u\", qp_id_base);\n+\tmbuf_mp = rte_pktmbuf_pool_create(mbuf_pool, nb_jobs * nb_qps, 0,\n \t\t\t0, MBUF_SIZE, rte_socket_id());\n \tif (mbuf_mp == NULL) {\n \t\tprintf(\"Error, can't create memory pool\\n\");\n@@ -402,7 +440,7 @@ run_regex(uint32_t nb_jobs,\n \t\t\t\t\t\tqp->total_enqueue +=\n \t\t\t\t\t\trte_regexdev_enqueue_burst\n \t\t\t\t\t\t\t(dev_id,\n-\t\t\t\t\t\t\tqp_id,\n+\t\t\t\t\t\t\tqp_id_base + qp_id,\n \t\t\t\t\t\t\tcur_ops_to_enqueue,\n \t\t\t\t\t\t\tactual_jobs -\n \t\t\t\t\t\t\tqp->total_enqueue);\n@@ -418,7 +456,7 @@ run_regex(uint32_t nb_jobs,\n \t\t\t\t\tqp->total_dequeue +=\n \t\t\t\t\t\trte_regexdev_dequeue_burst\n \t\t\t\t\t\t\t(dev_id,\n-\t\t\t\t\t\t\tqp_id,\n+\t\t\t\t\t\t\tqp_id_base + qp_id,\n \t\t\t\t\t\t\tcur_ops_to_dequeue,\n \t\t\t\t\t\t\tqp->total_enqueue -\n \t\t\t\t\t\t\tqp->total_dequeue);\n@@ -435,7 +473,7 @@ run_regex(uint32_t nb_jobs,\n \t (((double)actual_jobs * job_len * nb_iterations * 8) / time) /\n \t\t1000000000.0);\n \n-\tif (perf_mode)\n+\tif (rgxc->perf_mode)\n \t\tgoto end;\n \tfor (qp_id = 0; qp_id < nb_qps; qp_id++) {\n \t\tprintf(\"\\n############ QP id=%u ############\\n\", qp_id);\n@@ -491,6 +529,67 @@ run_regex(uint32_t nb_jobs,\n \treturn res;\n }\n \n+static int\n+distribute_qps_to_lcores(uint32_t nb_cores, uint32_t nb_qps,\n+\t\t\t struct qps_per_lcore **qpl)\n+{\n+\tint socket;\n+\tunsigned lcore_id;\n+\tuint32_t i;\n+\tuint16_t min_qp_id;\n+\tuint16_t max_qp_id;\n+\tstruct qps_per_lcore *qps_per_lcore;\n+\tuint32_t detected_lcores;\n+\n+\tif (nb_qps < nb_cores) {\n+\t\tnb_cores = nb_qps;\n+\t\tprintf(\"Reducing number of cores to number of QPs (%u)\\n\",\n+\t\t nb_cores);\n+\t}\n+\t/* Allocate qps_per_lcore array */\n+\tqps_per_lcore =\n+\t\trte_malloc(NULL, sizeof(*qps_per_lcore) * nb_cores, 0);\n+\tif (!qps_per_lcore)\n+\t\trte_exit(EXIT_FAILURE, \"Failed to create qps_per_lcore array\\n\");\n+\t*qpl = qps_per_lcore;\n+\tdetected_lcores = 0;\n+\tmin_qp_id = 0;\n+\n+\tRTE_LCORE_FOREACH_WORKER(lcore_id) {\n+\t\tif (detected_lcores >= nb_cores)\n+\t\t\tbreak;\n+\t\tqps_per_lcore[detected_lcores].lcore_id = lcore_id;\n+\t\tsocket = rte_lcore_to_socket_id(lcore_id);\n+\t\tif (socket == SOCKET_ID_ANY)\n+\t\t\tsocket = 0;\n+\t\tqps_per_lcore[detected_lcores].socket = socket;\n+\t\tqps_per_lcore[detected_lcores].qp_id_base = min_qp_id;\n+\t\tmax_qp_id = min_qp_id + nb_qps / nb_cores - 1;\n+\t\tif (nb_qps % nb_cores > detected_lcores)\n+\t\t\tmax_qp_id++;\n+\t\tqps_per_lcore[detected_lcores].nb_qps = max_qp_id -\n+\t\t\t\t\t\t\tmin_qp_id + 1;\n+\t\tmin_qp_id = max_qp_id + 1;\n+\t\tdetected_lcores++;\n+\t}\n+\tif (detected_lcores != nb_cores)\n+\t\treturn -1;\n+\n+\tfor (i = 0; i < detected_lcores; i++) {\n+\t\tprintf(\"===> Core %d: allocated queues: \",\n+\t\t qps_per_lcore[i].lcore_id);\n+\t\tmin_qp_id = qps_per_lcore[i].qp_id_base;\n+\t\tmax_qp_id =\n+\t\t\tqps_per_lcore[i].qp_id_base + qps_per_lcore[i].nb_qps;\n+\t\twhile (min_qp_id < max_qp_id) {\n+\t\t\tprintf(\"%u \", min_qp_id);\n+\t\t\tmin_qp_id++;\n+\t\t}\n+\t\tprintf(\"\\n\");\n+\t}\n+\treturn 0;\n+}\n+\n int\n main(int argc, char **argv)\n {\n@@ -506,6 +605,10 @@ main(int argc, char **argv)\n \tchar *data_buf;\n \tlong data_len;\n \tlong job_len;\n+\tuint32_t nb_lcores = 1;\n+\tstruct regex_conf *rgxc;\n+\tuint32_t i;\n+\tstruct qps_per_lcore *qps_per_lcore;\n \n \t/* Init EAL. */\n \tret = rte_eal_init(argc, argv);\n@@ -515,10 +618,15 @@ main(int argc, char **argv)\n \targv += ret;\n \tif (argc > 1)\n \t\targs_parse(argc, argv, rules_file, data_file, &nb_jobs,\n-\t\t\t\t&perf_mode, &nb_iterations, &nb_qps);\n+\t\t\t\t&perf_mode, &nb_iterations, &nb_qps,\n+\t\t\t\t&nb_lcores);\n \n \tif (nb_qps == 0)\n \t\trte_exit(EXIT_FAILURE, \"Number of QPs must be greater than 0\\n\");\n+\tif (nb_lcores == 0)\n+\t\trte_exit(EXIT_FAILURE, \"Number of lcores must be greater than 0\\n\");\n+\tif (distribute_qps_to_lcores(nb_lcores, nb_qps, &qps_per_lcore) < 0)\n+\t\trte_exit(EXIT_FAILURE, \"Failed to distribute queues to lcores!\\n\");\n \tret = init_port(&nb_max_payload, rules_file,\n \t\t\t&nb_max_matches, nb_qps);\n \tif (ret < 0)\n@@ -535,12 +643,27 @@ main(int argc, char **argv)\n \tif (job_len > nb_max_payload)\n \t\trte_exit(EXIT_FAILURE, \"Error, not enough jobs to cover input.\\n\");\n \n-\tret = run_regex(nb_jobs, perf_mode,\n-\t\t\tnb_iterations, nb_max_matches, nb_qps,\n-\t\t\tdata_buf, data_len, job_len);\n-\tif (ret < 0) {\n-\t\trte_exit(EXIT_FAILURE, \"RegEx function failed\\n\");\n+\trgxc = rte_malloc(NULL, sizeof(*rgxc) * nb_lcores, 0);\n+\tif (!rgxc)\n+\t\trte_exit(EXIT_FAILURE, \"Failed to create Regex Conf\\n\");\n+\tfor (i = 0; i < nb_lcores; i++) {\n+\t\trgxc[i] = (struct regex_conf){\n+\t\t\t.nb_jobs = nb_jobs,\n+\t\t\t.perf_mode = perf_mode,\n+\t\t\t.nb_iterations = nb_iterations,\n+\t\t\t.nb_max_matches = nb_max_matches,\n+\t\t\t.nb_qps = qps_per_lcore[i].nb_qps,\n+\t\t\t.qp_id_base = qps_per_lcore[i].qp_id_base,\n+\t\t\t.data_buf = data_buf,\n+\t\t\t.data_len = data_len,\n+\t\t\t.job_len = job_len,\n+\t\t};\n+\t\trte_eal_remote_launch(run_regex, &rgxc[i],\n+\t\t\t\t qps_per_lcore[i].lcore_id);\n \t}\n+\trte_eal_mp_wait_lcore();\n \trte_free(data_buf);\n+\trte_free(rgxc);\n+\trte_free(qps_per_lcore);\n \treturn EXIT_SUCCESS;\n }\ndiff --git a/doc/guides/tools/testregex.rst b/doc/guides/tools/testregex.rst\nindex 112b2bb..a59acd9 100644\n--- a/doc/guides/tools/testregex.rst\n+++ b/doc/guides/tools/testregex.rst\n@@ -7,13 +7,28 @@ dpdk-test-regex Tool\n The ``dpdk-test-regex`` tool is a Data Plane Development Kit (DPDK)\n application that allows functional testing and performance measurement for\n the RegEx PMDs.\n-The test supports only one core and one PMD.\n+\n It is based on precompiled rule file, and an input file, both of them can\n be selected using command-line options.\n \n In general case, each PMD has its own rule file.\n \n-The test outputs the following data:\n+By default the test supports one QP per core, however a higher number of cores\n+and QPs can be configured. The QPs are evenly distributed on the cores. All QPs\n+are assigned the same number of segments of input file to parse. Given n QPs\n+(per core) - the enqueue/dequeue RegEx operations are interleaved as follows::\n+\n+ enqueue(QP #1)\n+ enqueue(QP #2)\n+ ...\n+ enqueue(QP #n)\n+ dequeue(QP #1)\n+ dequeue(QP #2)\n+ ...\n+ dequeue(QP #n)\n+\n+\n+The test outputs the following data per QP and core:\n \n * Performance, in gigabit per second.\n \n@@ -26,8 +41,6 @@ The test outputs the following data:\n Limitations\n ~~~~~~~~~~~\n \n-* Only one queue is supported.\n-\n * Supports only precompiled rules.\n \n \n@@ -43,6 +56,12 @@ Application Options\n ``--nb_jobs N``\n number of jobs to use\n \n+``--nb_qps N``\n+ number of QPs to use\n+\n+``--nb_lcores N``\n+ number of cores to use\n+\n ``--perf N``\n only outputs the performance data\n \n@@ -70,4 +89,5 @@ The data file, will be used as a source data for the RegEx to work on.\n \n The tool has a number of command line options. Here is the sample command line::\n \n- ./dpdk-test-regex -a 83:00.0 -- --rules rule_file.rof2 --data data_file.txt --job 100\n+ ./dpdk-test-regex -a 83:00.0 -- --rules rule_file.rof2 --data data_file.txt --job 100 \\\n+ --nb_qps 4 --nb_lcores 2\n", "prefixes": [ "v3", "4/6" ] }