MINOR: pool: only use opportunistic versions of the swrate_add() functions
We don't need to know very accurately how much RAM is needed in a pool,
however we must not spend time competing with other threads trying to be
the one with the most accurate value. Let's use the "_opportunistic"
variants of swrate_add() which will simply cause some updates to be
dropped in case of thread contention. This should significantly improve
the situation when dealing with many threads and small per-thread caches.
Performance gains of up to 1-2% were observed on 48-thread systems thanks
to this alone.
diff --git a/src/pool.c b/src/pool.c
index df9d060..610e20c 100644
--- a/src/pool.c
+++ b/src/pool.c
@@ -380,7 +380,7 @@
if (!ptr)
return NULL;
- swrate_add_scaled(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used, POOL_AVG_SAMPLES/4);
+ swrate_add_scaled_opportunistic(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used, POOL_AVG_SAMPLES/4);
_HA_ATOMIC_INC(&pool->used);
/* keep track of where the element was allocated from */
@@ -396,7 +396,7 @@
void pool_free_nocache(struct pool_head *pool, void *ptr)
{
_HA_ATOMIC_DEC(&pool->used);
- swrate_add(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used);
+ swrate_add_opportunistic(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used);
pool_put_to_os(pool, ptr);
}
@@ -655,7 +655,7 @@
__ha_barrier_atomic_store();
} while (!_HA_ATOMIC_CAS(&pool->free_list, &free_list, item));
__ha_barrier_atomic_store();
- swrate_add(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used);
+ swrate_add_opportunistic(&pool->needed_avg, POOL_AVG_SAMPLES, pool->used);
}
/*