[PATCH v4 3/3] mempool: use cache for frequently updated stats

["Morten Brørup" <mb@smartsharesystems.com>]

When built with stats enabled (RTE_LIBRTE_MEMPOOL_STATS defined), the
performance of mempools with caches is improved as follows.

When accessing objects in the mempool, either the put_bulk and put_objs or
the get_success_bulk and get_success_objs statistics counters are likely
to be incremented.

By adding an alternative set of these counters to the mempool cache
structure, accessing the dedicated statistics structure is avoided in the
likely cases where these counters are incremented.

The trick here is that the cache line holding the mempool cache structure
is accessed anyway, in order to access the 'len' or 'flushthresh' fields.
Updating some statistics counters in the same cache line has lower
performance cost than accessing the statistics counters in the dedicated
statistics structure, which resides in another cache line.

mempool_perf_autotest with this patch shows the following improvements in
rate_persec.

The cost of enabling mempool stats (without debug) after this patch:
-6.8 % and -6.7 %, respectively without and with cache.

v4:
* Fix checkpatch warnings:
  A couple of typos in the patch description.
  The macro to add to a mempool cache stat variable should not use
  do {} while (0). Personally, I would tend to disagree with this, but
  whatever keeps the CI happy.
v3:
* Don't update the description of the RTE_MEMPOOL_STAT_ADD macro.
  This change belongs in the first patch of the series.
v2:
* Move the statistics counters into a stats structure.

Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
---
 lib/mempool/rte_mempool.c |  9 ++++++
 lib/mempool/rte_mempool.h | 66 ++++++++++++++++++++++++++++++++-------
 2 files changed, 64 insertions(+), 11 deletions(-)

diff --git a/lib/mempool/rte_mempool.c b/lib/mempool/rte_mempool.c
index e6208125e0..a18e39af04 100644
--- a/lib/mempool/rte_mempool.c
+++ b/lib/mempool/rte_mempool.c
@@ -1286,6 +1286,15 @@ rte_mempool_dump(FILE *f, struct rte_mempool *mp)
 		sum.get_success_blks += mp->stats[lcore_id].get_success_blks;
 		sum.get_fail_blks += mp->stats[lcore_id].get_fail_blks;
 	}
+	if (mp->cache_size != 0) {
+		/* Add the statistics stored in the mempool caches. */
+		for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
+			sum.put_bulk += mp->local_cache[lcore_id].stats.put_bulk;
+			sum.put_objs += mp->local_cache[lcore_id].stats.put_objs;
+			sum.get_success_bulk += mp->local_cache[lcore_id].stats.get_success_bulk;
+			sum.get_success_objs += mp->local_cache[lcore_id].stats.get_success_objs;
+		}
+	}
 	fprintf(f, "  stats:\n");
 	fprintf(f, "    put_bulk=%"PRIu64"\n", sum.put_bulk);
 	fprintf(f, "    put_objs=%"PRIu64"\n", sum.put_objs);
diff --git a/lib/mempool/rte_mempool.h b/lib/mempool/rte_mempool.h
index abfe34c05f..e6eb573739 100644
--- a/lib/mempool/rte_mempool.h
+++ b/lib/mempool/rte_mempool.h
@@ -86,6 +86,19 @@ struct rte_mempool_cache {
 	uint32_t size;	      /**< Size of the cache */
 	uint32_t flushthresh; /**< Threshold before we flush excess elements */
 	uint32_t len;	      /**< Current cache count */
+#ifdef RTE_LIBRTE_MEMPOOL_STATS
+	uint32_t unused;
+	/*
+	 * Alternative location for the most frequently updated mempool statistics (per-lcore),
+	 * providing faster update access when using a mempool cache.
+	 */
+	struct {
+		uint64_t put_bulk;          /**< Number of puts. */
+		uint64_t put_objs;          /**< Number of objects successfully put. */
+		uint64_t get_success_bulk;  /**< Successful allocation number. */
+		uint64_t get_success_objs;  /**< Objects successfully allocated. */
+	} stats;                        /**< Statistics */
+#endif
 	/**
 	 * Cache objects
 	 *
@@ -319,6 +332,22 @@ struct rte_mempool {
 #define RTE_MEMPOOL_STAT_ADD(mp, name, n) do {} while (0)
 #endif
 
+/**
+ * @internal When stats is enabled, store some statistics.
+ *
+ * @param cache
+ *   Pointer to the memory pool cache.
+ * @param name
+ *   Name of the statistics field to increment in the memory pool cache.
+ * @param n
+ *   Number to add to the statistics.
+ */
+#ifdef RTE_LIBRTE_MEMPOOL_STATS
+#define RTE_MEMPOOL_CACHE_STAT_ADD(cache, name, n) (cache)->stats.name += n
+#else
+#define RTE_MEMPOOL_CACHE_STAT_ADD(cache, name, n) do {} while (0)
+#endif
+
 /**
  * @internal Calculate the size of the mempool header.
  *
@@ -1333,13 +1362,17 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void * const *obj_table,
 {
 	void **cache_objs;
 
+	/* No cache provided */
+	if (unlikely(cache == NULL))
+		goto driver_enqueue;
+
 	/* increment stat now, adding in mempool always success */
-	RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
-	RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+	RTE_MEMPOOL_CACHE_STAT_ADD(cache, put_bulk, 1);
+	RTE_MEMPOOL_CACHE_STAT_ADD(cache, put_objs, n);
 
-	/* No cache provided or the request itself is too big for the cache */
-	if (unlikely(cache == NULL || n > cache->flushthresh))
-		goto driver_enqueue;
+	/* The request itself is too big for the cache */
+	if (unlikely(n > cache->flushthresh))
+		goto driver_enqueue_stats_incremented;
 
 	/*
 	 * The cache follows the following algorithm:
@@ -1364,6 +1397,12 @@ rte_mempool_do_generic_put(struct rte_mempool *mp, void * const *obj_table,
 
 driver_enqueue:
 
+	/* increment stat now, adding in mempool always success */
+	RTE_MEMPOOL_STAT_ADD(mp, put_bulk, 1);
+	RTE_MEMPOOL_STAT_ADD(mp, put_objs, n);
+
+driver_enqueue_stats_incremented:
+
 	/* push objects to the backend */
 	rte_mempool_ops_enqueue_bulk(mp, obj_table, n);
 }
@@ -1470,8 +1509,8 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void **obj_table,
 	if (remaining == 0) {
 		/* The entire request is satisfied from the cache. */
 
-		RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
-		RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+		RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_bulk, 1);
+		RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_objs, n);
 
 		return 0;
 	}
@@ -1500,8 +1539,8 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void **obj_table,
 
 	cache->len = cache->size;
 
-	RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
-	RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+	RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_bulk, 1);
+	RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_objs, n);
 
 	return 0;
 
@@ -1523,8 +1562,13 @@ rte_mempool_do_generic_get(struct rte_mempool *mp, void **obj_table,
 		RTE_MEMPOOL_STAT_ADD(mp, get_fail_bulk, 1);
 		RTE_MEMPOOL_STAT_ADD(mp, get_fail_objs, n);
 	} else {
-		RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
-		RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+		if (likely(cache != NULL)) {
+			RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_bulk, 1);
+			RTE_MEMPOOL_CACHE_STAT_ADD(cache, get_success_objs, n);
+		} else {
+			RTE_MEMPOOL_STAT_ADD(mp, get_success_bulk, 1);
+			RTE_MEMPOOL_STAT_ADD(mp, get_success_objs, n);
+		}
 	}
 
 	return ret;
-- 
2.17.1