[dpdk-dev] [PATCH] ring: empty and count optimizations
Ananyev, Konstantin
konstantin.ananyev at intel.com
Thu May 14 14:23:31 CEST 2020
Hi Morten,
> Testing if the ring is empty is as simple as comparing the producer and
> consumer pointers.
> In theory, this optimization reduces the number of potential cache misses
> from 3 to 2 by not having to read r->mask in rte_ring_count().
>
> It is not possible to enqueue more elements than the capacity of a ring,
> so the capacity comparison is a safeguard for observer threads only.
> Instead of completely removing the comparison, I have reorganized it to
> resemble the other trigrahps in the ring library and added a likely().
>
> The modification of these two functions were discussed in the RFC here:
> https://mails.dpdk.org/archives/dev/2020-April/165752.html
>
> Also fixed some existing code not passing checkpatch.
>
> Signed-off-by: Morten Brørup <mb at smartsharesystems.com>
> ---
> lib/librte_ring/rte_ring.h | 36 +++++++++++++++++++-----------------
> 1 file changed, 19 insertions(+), 17 deletions(-)
>
> diff --git a/lib/librte_ring/rte_ring.h b/lib/librte_ring/rte_ring.h
> index 86faede81..36438d9cd 100644
> --- a/lib/librte_ring/rte_ring.h
> +++ b/lib/librte_ring/rte_ring.h
> @@ -55,7 +55,7 @@ extern "C" {
> * - The memory size needed for the ring on success.
> * - -EINVAL if count is not a power of 2.
> */
> -ssize_t rte_ring_get_memsize(unsigned count);
> +ssize_t rte_ring_get_memsize(unsigned int count);
All these changes to replace 'unsigned' with insigned int' -
seems to be irrelevant to the patch subject, so can you
put them to a separate patch in the series.
> /**
> * Initialize a ring structure.
> @@ -109,8 +109,8 @@ ssize_t rte_ring_get_memsize(unsigned count);
> * @return
> * 0 on success, or a negative value on error.
> */
> -int rte_ring_init(struct rte_ring *r, const char *name, unsigned count,
> - unsigned flags);
> +int rte_ring_init(struct rte_ring *r, const char *name, unsigned int count,
> + unsigned int flags);
>
> /**
> * Create a new ring named *name* in memory.
> @@ -169,8 +169,8 @@ int rte_ring_init(struct rte_ring *r, const char *name, unsigned count,
> * - EEXIST - a memzone with the same name already exists
> * - ENOMEM - no appropriate memory area found in which to create memzone
> */
> -struct rte_ring *rte_ring_create(const char *name, unsigned count,
> - int socket_id, unsigned flags);
> +struct rte_ring *rte_ring_create(const char *name, unsigned int count,
> + int socket_id, unsigned int flags);
>
> /**
> * De-allocate all memory used by the ring.
> @@ -199,7 +199,7 @@ void rte_ring_dump(FILE *f, const struct rte_ring *r);
> uint32_t idx = prod_head & (r)->mask; \
> obj_type *ring = (obj_type *)ring_start; \
> if (likely(idx + n < size)) { \
> - for (i = 0; i < (n & ((~(unsigned)0x3))); i+=4, idx+=4) { \
> + for (i = 0; i < (n & ~0x3); i += 4, idx += 4) { \
> ring[idx] = obj_table[i]; \
> ring[idx+1] = obj_table[i+1]; \
> ring[idx+2] = obj_table[i+2]; \
> @@ -230,7 +230,7 @@ void rte_ring_dump(FILE *f, const struct rte_ring *r);
> const uint32_t size = (r)->size; \
> obj_type *ring = (obj_type *)ring_start; \
> if (likely(idx + n < size)) { \
> - for (i = 0; i < (n & (~(unsigned)0x3)); i+=4, idx+=4) {\
> + for (i = 0; i < (n & ~0x3); i += 4, idx += 4) {\
> obj_table[i] = ring[idx]; \
> obj_table[i+1] = ring[idx+1]; \
> obj_table[i+2] = ring[idx+2]; \
> @@ -683,13 +683,13 @@ rte_ring_reset(struct rte_ring *r);
> * @return
> * The number of entries in the ring.
> */
> -static inline unsigned
> +static inline unsigned int
> rte_ring_count(const struct rte_ring *r)
> {
> uint32_t prod_tail = r->prod.tail;
> uint32_t cons_tail = r->cons.tail;
> uint32_t count = (prod_tail - cons_tail) & r->mask;
> - return (count > r->capacity) ? r->capacity : count;
> + return likely(count <= r->capacity) ? count : r->capacity;
Honestly, I don't see there is any point of that change:
I think it wouldn't change anything in terms of functionality
or performance.
> }
>
> /**
> @@ -700,7 +700,7 @@ rte_ring_count(const struct rte_ring *r)
> * @return
> * The number of free entries in the ring.
> */
> -static inline unsigned
> +static inline unsigned int
> rte_ring_free_count(const struct rte_ring *r)
> {
> return r->capacity - rte_ring_count(r);
> @@ -733,7 +733,9 @@ rte_ring_full(const struct rte_ring *r)
> static inline int
> rte_ring_empty(const struct rte_ring *r)
> {
> - return rte_ring_count(r) == 0;
> + uint32_t prod_tail = r->prod.tail;
> + uint32_t cons_tail = r->cons.tail;
> + return cons_tail == prod_tail;
> }
>
> /**
> @@ -860,7 +862,7 @@ struct rte_ring *rte_ring_lookup(const char *name);
> * @return
> * - n: Actual number of objects enqueued.
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_mp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> unsigned int n, unsigned int *free_space)
> {
> @@ -883,7 +885,7 @@ rte_ring_mp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> * @return
> * - n: Actual number of objects enqueued.
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_sp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> unsigned int n, unsigned int *free_space)
> {
> @@ -910,7 +912,7 @@ rte_ring_sp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> * @return
> * - n: Actual number of objects enqueued.
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> unsigned int n, unsigned int *free_space)
> {
> @@ -954,7 +956,7 @@ rte_ring_enqueue_burst(struct rte_ring *r, void * const *obj_table,
> * @return
> * - n: Actual number of objects dequeued, 0 if ring is empty
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_mc_dequeue_burst(struct rte_ring *r, void **obj_table,
> unsigned int n, unsigned int *available)
> {
> @@ -979,7 +981,7 @@ rte_ring_mc_dequeue_burst(struct rte_ring *r, void **obj_table,
> * @return
> * - n: Actual number of objects dequeued, 0 if ring is empty
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_sc_dequeue_burst(struct rte_ring *r, void **obj_table,
> unsigned int n, unsigned int *available)
> {
> @@ -1006,7 +1008,7 @@ rte_ring_sc_dequeue_burst(struct rte_ring *r, void **obj_table,
> * @return
> * - Number of objects dequeued
> */
> -static __rte_always_inline unsigned
> +static __rte_always_inline unsigned int
> rte_ring_dequeue_burst(struct rte_ring *r, void **obj_table,
> unsigned int n, unsigned int *available)
> {
> --
> 2.17.1
More information about the dev
mailing list