From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 2BC1A42C61; Thu, 8 Jun 2023 17:26:53 +0200 (CEST) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 104EA42D20; Thu, 8 Jun 2023 17:26:20 +0200 (CEST) Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by mails.dpdk.org (Postfix) with ESMTP id 3118942D51 for ; Thu, 8 Jun 2023 17:26:18 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1686237978; x=1717773978; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=LEUCKUP7up/Uc/vD00veRF3W1Nw0GJubk4khQlc5XI4=; b=E8M5zloBBm5FrtQa4uXyJErLTaaL8raawVhaVgaFTdo2eSZFbK8r2Nmx 1gwQnnCvAgfKQ8ccZiRuUfYyD3a528yGWFjHsXcbGTshVANj2H/uR5sdY raY5n0lg+oFbbAvcmjwzSNBa93pKLZL5FhHFsaGePCehAIEtsXKMXVBQi lTiW7m9CiH4LVqZ7MSEQtt7I1XLNv0BwpT2sZagA/v95vXFRl844dxFQB 9pR9/nno7QEKSyO4gEKFYX8Krs4S6rLTtbIYghZQnrWJ4kXajAd1dxQiC 6zBknBL+/rI3DT3oraCWjNicn3cbqh77u/diqWM/WPf+Gxt3VpRiiyq/7 A==; X-IronPort-AV: E=McAfee;i="6600,9927,10735"; a="385681496" X-IronPort-AV: E=Sophos;i="6.00,227,1681196400"; d="scan'208";a="385681496" Received: from orsmga004.jf.intel.com ([10.7.209.38]) by fmsmga101.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 08 Jun 2023 08:26:17 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10735"; a="834233064" X-IronPort-AV: E=Sophos;i="6.00,227,1681196400"; d="scan'208";a="834233064" Received: from dpdk-zhirun-lmm.sh.intel.com ([10.67.119.94]) by orsmga004.jf.intel.com with ESMTP; 08 Jun 2023 08:26:13 -0700 From: Zhirun Yan To: dev@dpdk.org, jerinj@marvell.com, kirankumark@marvell.com, ndabilpuram@marvell.com, stephen@networkplumber.org, pbhagavatula@marvell.com, jerinjacobk@gmail.com Cc: cunming.liang@intel.com, haiyue.wang@intel.com, mattias.ronnblom@ericsson.com, Zhirun Yan Subject: [PATCH v11 10/16] graph: introduce stream moving cross cores Date: Thu, 8 Jun 2023 23:18:38 +0800 Message-Id: <20230608151844.1823783-11-zhirun.yan@intel.com> X-Mailer: git-send-email 2.37.2 In-Reply-To: <20230608151844.1823783-1-zhirun.yan@intel.com> References: <20230608095759.1800617-1-zhirun.yan@intel.com> <20230608151844.1823783-1-zhirun.yan@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org This patch introduces key functions to allow a worker thread to enable enqueue and move streams of objects to the next nodes over different cores for mcore dispatch model. Signed-off-by: Haiyue Wang Signed-off-by: Cunming Liang Signed-off-by: Zhirun Yan Acked-by: Pavan Nikhilesh --- lib/graph/graph.c | 6 +- lib/graph/graph_private.h | 31 ++++ lib/graph/meson.build | 2 +- lib/graph/rte_graph.h | 15 +- lib/graph/rte_graph_model_mcore_dispatch.c | 158 +++++++++++++++++++++ lib/graph/rte_graph_model_mcore_dispatch.h | 45 ++++++ lib/graph/version.map | 3 + 7 files changed, 255 insertions(+), 5 deletions(-) diff --git a/lib/graph/graph.c b/lib/graph/graph.c index 968cbbf86c..41251e3435 100644 --- a/lib/graph/graph.c +++ b/lib/graph/graph.c @@ -473,7 +473,7 @@ rte_graph_destroy(rte_graph_t id) } static rte_graph_t -graph_clone(struct graph *parent_graph, const char *name) +graph_clone(struct graph *parent_graph, const char *name, struct rte_graph_param *prm) { struct graph_node *graph_node; struct graph *graph; @@ -547,14 +547,14 @@ graph_clone(struct graph *parent_graph, const char *name) } rte_graph_t -rte_graph_clone(rte_graph_t id, const char *name) +rte_graph_clone(rte_graph_t id, const char *name, struct rte_graph_param *prm) { struct graph *graph; GRAPH_ID_CHECK(id); STAILQ_FOREACH(graph, &graph_list, next) if (graph->id == id) - return graph_clone(graph, name); + return graph_clone(graph, name, prm); fail: return RTE_GRAPH_ID_INVALID; diff --git a/lib/graph/graph_private.h b/lib/graph/graph_private.h index d84174b667..d0ef13b205 100644 --- a/lib/graph/graph_private.h +++ b/lib/graph/graph_private.h @@ -414,4 +414,35 @@ void graph_dump(FILE *f, struct graph *g); */ void node_dump(FILE *f, struct node *n); +/** + * @internal + * + * Create the graph schedule work queue for mcore dispatch model. + * All cloned graphs attached to the parent graph MUST be destroyed together + * for fast schedule design limitation. + * + * @param _graph + * The graph object + * @param _parent_graph + * The parent graph object which holds the run-queue head. + * @param prm + * Graph parameter, includes model-specific parameters in this graph. + * + * @return + * - 0: Success. + * - <0: Graph schedule work queue related error. + */ +int graph_sched_wq_create(struct graph *_graph, struct graph *_parent_graph, + struct rte_graph_param *prm); + +/** + * @internal + * + * Destroy the graph schedule work queue for mcore dispatch model. + * + * @param _graph + * The graph object + */ +void graph_sched_wq_destroy(struct graph *_graph); + #endif /* _RTE_GRAPH_PRIVATE_H_ */ diff --git a/lib/graph/meson.build b/lib/graph/meson.build index 0685cf9e72..9d51eabe33 100644 --- a/lib/graph/meson.build +++ b/lib/graph/meson.build @@ -20,4 +20,4 @@ sources = files( ) headers = files('rte_graph.h', 'rte_graph_worker.h') -deps += ['eal', 'pcapng'] +deps += ['eal', 'pcapng', 'mempool', 'ring'] diff --git a/lib/graph/rte_graph.h b/lib/graph/rte_graph.h index 998cade200..2ffee520b1 100644 --- a/lib/graph/rte_graph.h +++ b/lib/graph/rte_graph.h @@ -169,6 +169,17 @@ struct rte_graph_param { bool pcap_enable; /**< Pcap enable. */ uint64_t num_pkt_to_capture; /**< Number of packets to capture. */ char *pcap_filename; /**< Filename in which packets to be captured.*/ + + RTE_STD_C11 + union { + struct { + uint64_t rsvd; /**< Reserved for rtc model. */ + } rtc; + struct { + uint32_t wq_size_max; /**< Maximum size of workqueue for dispatch model. */ + uint32_t mp_capacity; /**< Capacity of memory pool for dispatch model. */ + } dispatch; + }; }; /** @@ -260,12 +271,14 @@ int rte_graph_destroy(rte_graph_t id); * Name of the new graph. The library prepends the parent graph name to the * user-specified name. The final graph name will be, * "parent graph name" + "-" + name. + * @param prm + * Graph parameter, includes model-specific parameters in this graph. * * @return * Valid graph id on success, RTE_GRAPH_ID_INVALID otherwise. */ __rte_experimental -rte_graph_t rte_graph_clone(rte_graph_t id, const char *name); +rte_graph_t rte_graph_clone(rte_graph_t id, const char *name, struct rte_graph_param *prm); /** * Get graph id from graph name. diff --git a/lib/graph/rte_graph_model_mcore_dispatch.c b/lib/graph/rte_graph_model_mcore_dispatch.c index 9df2479a10..8f4bc860ab 100644 --- a/lib/graph/rte_graph_model_mcore_dispatch.c +++ b/lib/graph/rte_graph_model_mcore_dispatch.c @@ -5,6 +5,164 @@ #include "graph_private.h" #include "rte_graph_model_mcore_dispatch.h" +int +graph_sched_wq_create(struct graph *_graph, struct graph *_parent_graph, + struct rte_graph_param *prm) +{ + struct rte_graph *parent_graph = _parent_graph->graph; + struct rte_graph *graph = _graph->graph; + unsigned int wq_size; + unsigned int flags = RING_F_SC_DEQ; + + wq_size = GRAPH_SCHED_WQ_SIZE(graph->nb_nodes); + wq_size = rte_align32pow2(wq_size + 1); + + if (prm->dispatch.wq_size_max > 0) + wq_size = wq_size <= (prm->dispatch.wq_size_max) ? wq_size : + prm->dispatch.wq_size_max; + + if (!rte_is_power_of_2(wq_size)) + flags |= RING_F_EXACT_SZ; + + graph->dispatch.wq = rte_ring_create(graph->name, wq_size, graph->socket, + flags); + if (graph->dispatch.wq == NULL) + SET_ERR_JMP(EIO, fail, "Failed to allocate graph WQ"); + + if (prm->dispatch.mp_capacity > 0) + wq_size = (wq_size <= prm->dispatch.mp_capacity) ? wq_size : + prm->dispatch.mp_capacity; + + graph->dispatch.mp = rte_mempool_create(graph->name, wq_size, + sizeof(struct graph_mcore_dispatch_wq_node), + 0, 0, NULL, NULL, NULL, NULL, + graph->socket, MEMPOOL_F_SP_PUT); + if (graph->dispatch.mp == NULL) + SET_ERR_JMP(EIO, fail_mp, + "Failed to allocate graph WQ schedule entry"); + + graph->dispatch.lcore_id = _graph->lcore_id; + + if (parent_graph->dispatch.rq == NULL) { + parent_graph->dispatch.rq = &parent_graph->dispatch.rq_head; + SLIST_INIT(parent_graph->dispatch.rq); + } + + graph->dispatch.rq = parent_graph->dispatch.rq; + SLIST_INSERT_HEAD(graph->dispatch.rq, graph, next); + + return 0; + +fail_mp: + rte_ring_free(graph->dispatch.wq); + graph->dispatch.wq = NULL; +fail: + return -rte_errno; +} + +void +graph_sched_wq_destroy(struct graph *_graph) +{ + struct rte_graph *graph = _graph->graph; + + if (graph == NULL) + return; + + rte_ring_free(graph->dispatch.wq); + graph->dispatch.wq = NULL; + + rte_mempool_free(graph->dispatch.mp); + graph->dispatch.mp = NULL; +} + +static __rte_always_inline bool +__graph_sched_node_enqueue(struct rte_node *node, struct rte_graph *graph) +{ + struct graph_mcore_dispatch_wq_node *wq_node; + uint16_t off = 0; + uint16_t size; + +submit_again: + if (rte_mempool_get(graph->dispatch.mp, (void **)&wq_node) < 0) + goto fallback; + + size = RTE_MIN(node->idx, RTE_DIM(wq_node->objs)); + wq_node->node_off = node->off; + wq_node->nb_objs = size; + rte_memcpy(wq_node->objs, &node->objs[off], size * sizeof(void *)); + + while (rte_ring_mp_enqueue_bulk_elem(graph->dispatch.wq, (void *)&wq_node, + sizeof(wq_node), 1, NULL) == 0) + rte_pause(); + + off += size; + node->idx -= size; + if (node->idx > 0) + goto submit_again; + + return true; + +fallback: + if (off != 0) + memmove(&node->objs[0], &node->objs[off], + node->idx * sizeof(void *)); + + return false; +} + +bool __rte_noinline +__rte_graph_mcore_dispatch_sched_node_enqueue(struct rte_node *node, + struct rte_graph_rq_head *rq) +{ + const unsigned int lcore_id = node->dispatch.lcore_id; + struct rte_graph *graph; + + SLIST_FOREACH(graph, rq, next) + if (graph->dispatch.lcore_id == lcore_id) + break; + + return graph != NULL ? __graph_sched_node_enqueue(node, graph) : false; +} + +void +__rte_graph_mcore_dispatch_sched_wq_process(struct rte_graph *graph) +{ +#define WQ_SZ 32 + struct graph_mcore_dispatch_wq_node *wq_node; + struct rte_mempool *mp = graph->dispatch.mp; + struct rte_ring *wq = graph->dispatch.wq; + uint16_t idx, free_space; + struct rte_node *node; + unsigned int i, n; + struct graph_mcore_dispatch_wq_node *wq_nodes[WQ_SZ]; + + n = rte_ring_sc_dequeue_burst_elem(wq, wq_nodes, sizeof(wq_nodes[0]), + RTE_DIM(wq_nodes), NULL); + if (n == 0) + return; + + for (i = 0; i < n; i++) { + wq_node = wq_nodes[i]; + node = RTE_PTR_ADD(graph, wq_node->node_off); + RTE_ASSERT(node->fence == RTE_GRAPH_FENCE); + idx = node->idx; + free_space = node->size - idx; + + if (unlikely(free_space < wq_node->nb_objs)) + __rte_node_stream_alloc_size(graph, node, node->size + wq_node->nb_objs); + + memmove(&node->objs[idx], wq_node->objs, wq_node->nb_objs * sizeof(void *)); + node->idx = idx + wq_node->nb_objs; + + __rte_node_process(graph, node); + + wq_node->nb_objs = 0; + node->idx = 0; + } + + rte_mempool_put_bulk(mp, (void **)wq_nodes, n); +} + int rte_graph_model_mcore_dispatch_node_lcore_affinity_set(const char *name, unsigned int lcore_id) { diff --git a/lib/graph/rte_graph_model_mcore_dispatch.h b/lib/graph/rte_graph_model_mcore_dispatch.h index 7da0483d13..6163f96c37 100644 --- a/lib/graph/rte_graph_model_mcore_dispatch.h +++ b/lib/graph/rte_graph_model_mcore_dispatch.h @@ -20,8 +20,53 @@ extern "C" { #endif +#include +#include +#include +#include + #include "rte_graph_worker_common.h" +#define GRAPH_SCHED_WQ_SIZE_MULTIPLIER 8 +#define GRAPH_SCHED_WQ_SIZE(nb_nodes) \ + ((typeof(nb_nodes))((nb_nodes) * GRAPH_SCHED_WQ_SIZE_MULTIPLIER)) + +/** + * @internal + * + * Schedule the node to the right graph's work queue for mcore dispatch model. + * + * @param node + * Pointer to the scheduled node object. + * @param rq + * Pointer to the scheduled run-queue for all graphs. + * + * @return + * True on success, false otherwise. + * + * @note + * This implementation is used by mcore dispatch model only and user application + * should not call it directly. + */ +__rte_experimental +bool __rte_noinline __rte_graph_mcore_dispatch_sched_node_enqueue(struct rte_node *node, + struct rte_graph_rq_head *rq); + +/** + * @internal + * + * Process all nodes (streams) in the graph's work queue for mcore dispatch model. + * + * @param graph + * Pointer to the graph object. + * + * @note + * This implementation is used by mcore dispatch model only and user application + * should not call it directly. + */ +__rte_experimental +void __rte_graph_mcore_dispatch_sched_wq_process(struct rte_graph *graph); + /** * Set lcore affinity with the node used for mcore dispatch model. * diff --git a/lib/graph/version.map b/lib/graph/version.map index 9e92b54ffa..7e985d6308 100644 --- a/lib/graph/version.map +++ b/lib/graph/version.map @@ -1,6 +1,9 @@ EXPERIMENTAL { global: + __rte_graph_mcore_dispatch_sched_node_enqueue; + __rte_graph_mcore_dispatch_sched_wq_process; + __rte_node_register; __rte_node_stream_alloc; __rte_node_stream_alloc_size; -- 2.37.2