[PATCH v6 09/15] graph: introduce stream moving cross cores

[Jerin Jacob]

On Tue, May 9, 2023 at 11:35 AM Zhirun Yan <zhirun.yan at intel.com> wrote:
>
> 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.

different cores-> different cores for mcore dispatch model.


>
> Signed-off-by: Haiyue Wang <haiyue.wang at intel.com>
> Signed-off-by: Cunming Liang <cunming.liang at intel.com>
> Signed-off-by: Zhirun Yan <zhirun.yan at intel.com>
> ---
>  lib/graph/graph.c                    |   6 +-
>  lib/graph/graph_private.h            |  30 +++++
>  lib/graph/meson.build                |   2 +-
>  lib/graph/rte_graph.h                |  15 ++-
>  lib/graph/rte_graph_model_dispatch.c | 157 +++++++++++++++++++++++++++
>  lib/graph/rte_graph_model_dispatch.h |  37 +++++++
>  lib/graph/version.map                |   2 +
>  7 files changed, 244 insertions(+), 5 deletions(-)
>
> diff --git a/lib/graph/graph.c b/lib/graph/graph.c
> index e809aa55b0..f555844d8f 100644
> --- a/lib/graph/graph.c
> +++ b/lib/graph/graph.c
> @@ -495,7 +495,7 @@ clone_name(struct graph *graph, struct graph *parent_graph, const char *name)
>  }
>
>  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;
> @@ -566,14 +566,14 @@ graph_clone(struct graph *parent_graph, const char *name)
>  }
>
> --- 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[8];
> +               } rtc;
> +               struct {
> +                       uint32_t wq_size_max;
> +                       uint32_t mp_capacity;

Add doxgen comment for all please.

> +               } 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);
>
>  /**
> +void
> +__rte_graph_sched_wq_process(struct rte_graph *graph)
> +{
> +       struct graph_sched_wq_node *wq_node;
> +       struct rte_mempool *mp = graph->mp;
> +       struct rte_ring *wq = graph->wq;
> +       uint16_t idx, free_space;
> +       struct rte_node *node;
> +       unsigned int i, n;
> +       struct graph_sched_wq_node *wq_nodes[32];

Use RTE_GRAPH_BURST_SIZE instead of 32, if it is anything do with
burst size? else ignore.


> +
> +       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);
> +}
> +
> +/**
> + * @internal

For both internal function, you can add Doxygen comment as @note to
tell this must not be used directly.

> + *
> + * Process all nodes (streams) in the graph's work queue.
> + *
> + * @param graph
> + *   Pointer to the graph object.
> + */
> +__rte_experimental
> +void __rte_graph_sched_wq_process(struct rte_graph *graph);
> +
>  /**
>   * Set lcore affinity with the node.
>   *
> diff --git a/lib/graph/version.map b/lib/graph/version.map
> index aaa86f66ed..d511133f39 100644
> --- a/lib/graph/version.map
> +++ b/lib/graph/version.map
> @@ -48,6 +48,8 @@ EXPERIMENTAL {
>
>         rte_graph_worker_model_set;
>         rte_graph_worker_model_get;
> +       __rte_graph_sched_wq_process;
> +       __rte_graph_sched_node_enqueue;

Please add _mcore_dispatch_ name space.



>
>         rte_graph_model_dispatch_lcore_affinity_set;
>
> --
> 2.37.2
>