On Tue, Oct 15, 2024 at 7:43 AM <pbhagavat...@marvell.com> wrote:
@@ -88,4 +90,10 @@ node_mbuf_priv2(struct rte_mbuf *m)
> return (struct node_mbuf_priv2 *)rte_mbuf_to_priv(m);
> }
>
> +#define NODE_INCREMENT_XSTAT_ID(node, id, cond, cnt)
> \
> + {
> \
> + if (unlikely(rte_graph_has_stats_feature() && (cond)))
> \
> + ((uint64_t *)RTE_PTR_ADD(node, node->xstat_off))[id]
> += (cnt); \
> + }
> +
> #endif /* __NODE_PRIVATE_H__ */
Why introduce rte_node_xstat_increment() in patch 1, but not use it here?
I'll merge this as is, but recoding such helper as a macro in node
code is going in a wrong direction.
--
David Marchand
