Added cache guard after the table holding the ring elements, to avoid false sharing conflicts caused by next-line hardware prefetchers when accessing elements at the end of the ring table.
Signed-off-by: Morten Brørup <[email protected]> Acked-by: Konstantin Ananyev <[email protected]> --- v2: * Added comment describing reason for padding. (Konstantin) --- lib/ring/rte_ring.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/lib/ring/rte_ring.c b/lib/ring/rte_ring.c index f10050a1c4..10b52dc679 100644 --- a/lib/ring/rte_ring.c +++ b/lib/ring/rte_ring.c @@ -73,8 +73,15 @@ rte_ring_get_memsize_elem(unsigned int esize, unsigned int count) return -EINVAL; } + static_assert(sizeof(struct rte_ring) == RTE_CACHE_LINE_ROUNDUP(sizeof(struct rte_ring)), + "Size of struct rte_ring not cache line aligned"); sz = sizeof(struct rte_ring) + (ssize_t)count * esize; + /* Add padding, to guard against false sharing-like effects + * on systems with a next-N-lines hardware prefetcher, when + * accessing elements at the end of the ring table. + */ sz = RTE_ALIGN(sz, RTE_CACHE_LINE_SIZE); + sz += RTE_CACHE_GUARD_LINES * RTE_CACHE_LINE_SIZE; return sz; } -- 2.43.0
