From: "He, Jia" <jia.he@hxt-semitech.com>
To: Gavin Hu <gavin.hu@arm.com>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "Honnappa.Nagarahalli@arm.com" <Honnappa.Nagarahalli@arm.com>,
"steve.capper@arm.com" <steve.capper@arm.com>,
"Ola.Liljedahl@arm.com" <Ola.Liljedahl@arm.com>,
"jerin.jacob@caviumnetworks.com" <jerin.jacob@caviumnetworks.com>,
"hemant.agrawal@nxp.com" <hemant.agrawal@nxp.com>,
"stable@dpdk.org" <stable@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v2] ring: fix c11 memory ordering issue
Date: Tue, 7 Aug 2018 05:56:36 +0000 [thread overview]
Message-ID: <88b4421b63ec4802892f3fc40d457f03@HXTBJIDCEMVIW01.hxtcorp.net> (raw)
In-Reply-To: <20180807031943.5331-1-gavin.hu@arm.com>
Hi Gavin
> -----Original Message-----
> From: Gavin Hu [mailto:gavin.hu@arm.com]
> Sent: 2018年8月7日 11:20
> To: dev@dpdk.org
> Cc: gavin.hu@arm.com; Honnappa.Nagarahalli@arm.com;
> steve.capper@arm.com; Ola.Liljedahl@arm.com;
> jerin.jacob@caviumnetworks.com; hemant.agrawal@nxp.com; He, Jia
> <jia.he@hxt-semitech.com>; stable@dpdk.org
> Subject: [PATCH v2] ring: fix c11 memory ordering issue
>
> This patch includes two bug fixes(#1 and 2) and two optimisations(#3 and #4).
Maybe you need to split this into small parts.
> 1) In update_tail, read ht->tail using __atomic_load.Although the
> compiler currently seems to be doing the right thing even without
> _atomic_load, we don't want to give the compiler freedom to optimise
> what should be an atomic load, it should not be arbitarily moved
> around.
> 2) Synchronize the load-acquire of the tail and the store-release
> within update_tail, the store release ensures all the ring operations,
> engqueu or dequeue are seen by the observers as soon as they see
> the updated tail. The load-acquire is required for correctly compu-
> tate the free_entries or avail_entries, respectively for enqueue and
> dequeue operations, the data dependency is not reliable for ordering
> as the compiler might break it by saving to temporary values to boost
> performance.
Could you describe the race condition in details?
e.g.
cpu 1 cpu2
code1
code2
Cheers,
Jia
> 3) In __rte_ring_move_prod_head, move the __atomic_load_n up and out of
> the do {} while loop as upon failure the old_head will be updated,
> another load is costy and not necessary.
> 4) When calling __atomic_compare_exchange_n, relaxed ordering for both
> success and failure, as multiple threads can work independently on
> the same end of the ring (either enqueue or dequeue) without
> synchronization, not as operating on tail, which has to be finished
> in sequence.
>
> The patch was benchmarked with test/ring_perf_autotest and it decreases the
> enqueue/dequeue latency by 5% ~ 24.6% with two lcores, the real gains are
> dependent on the number of lcores, depth of the ring, SPSC or MPMC.
> For 1 lcore, it also improves a little, about 3 ~ 4%.
> It is a big improvement, in case of MPMC, with rings size of 32, it saves latency up
> to (6.90-5.20)/6.90 = 24.6%.
>
> Test result data:
>
> SP/SC bulk enq/dequeue (size: 8): 13.19
> MP/MC bulk enq/dequeue (size: 8): 25.79
> SP/SC bulk enq/dequeue (size: 32): 3.85
> MP/MC bulk enq/dequeue (size: 32): 6.90
>
> SP/SC bulk enq/dequeue (size: 8): 12.05
> MP/MC bulk enq/dequeue (size: 8): 23.06
> SP/SC bulk enq/dequeue (size: 32): 3.62
> MP/MC bulk enq/dequeue (size: 32): 5.20
>
> Fixes: 39368ebfc6 ("ring: introduce C11 memory model barrier option")
> Cc: stable@dpdk.org
>
> Signed-off-by: Gavin Hu <gavin.hu@arm.com>
> Reviewed-by: Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com>
> Reviewed-by: Steve Capper <steve.capper@arm.com>
> Reviewed-by: Ola Liljedahl <Ola.Liljedahl@arm.com>
> ---
> lib/librte_ring/rte_ring_c11_mem.h | 38
> +++++++++++++++++++++++++-------------
> 1 file changed, 25 insertions(+), 13 deletions(-)
>
> diff --git a/lib/librte_ring/rte_ring_c11_mem.h
> b/lib/librte_ring/rte_ring_c11_mem.h
> index 94df3c4a6..cfa3be4a7 100644
> --- a/lib/librte_ring/rte_ring_c11_mem.h
> +++ b/lib/librte_ring/rte_ring_c11_mem.h
> @@ -21,7 +21,8 @@ update_tail(struct rte_ring_headtail *ht, uint32_t old_val,
> uint32_t new_val,
> * we need to wait for them to complete
> */
> if (!single)
> - while (unlikely(ht->tail != old_val))
> + while (unlikely(old_val != __atomic_load_n(&ht->tail,
> + __ATOMIC_RELAXED)))
> rte_pause();
>
> __atomic_store_n(&ht->tail, new_val, __ATOMIC_RELEASE); @@ -60,20
> +61,24 @@ __rte_ring_move_prod_head(struct rte_ring *r, unsigned int is_sp,
> unsigned int max = n;
> int success;
>
> + *old_head = __atomic_load_n(&r->prod.head, __ATOMIC_RELAXED);
> do {
> /* Reset n to the initial burst count */
> n = max;
>
> - *old_head = __atomic_load_n(&r->prod.head,
> - __ATOMIC_ACQUIRE);
>
> - /*
> - * The subtraction is done between two unsigned 32bits value
> + /* load-acquire synchronize with store-release of ht->tail
> + * in update_tail.
> + */
> + const uint32_t cons_tail = __atomic_load_n(&r->cons.tail,
> + __ATOMIC_ACQUIRE);
> +
> + /* The subtraction is done between two unsigned 32bits value
> * (the result is always modulo 32 bits even if we have
> * *old_head > cons_tail). So 'free_entries' is always between 0
> * and capacity (which is < size).
> */
> - *free_entries = (capacity + r->cons.tail - *old_head);
> + *free_entries = (capacity + cons_tail - *old_head);
>
> /* check that we have enough room in ring */
> if (unlikely(n > *free_entries))
> @@ -87,9 +92,10 @@ __rte_ring_move_prod_head(struct rte_ring *r, unsigned
> int is_sp,
> if (is_sp)
> r->prod.head = *new_head, success = 1;
> else
> + /* on failure, *old_head is updated */
> success = __atomic_compare_exchange_n(&r->prod.head,
> old_head, *new_head,
> - 0, __ATOMIC_ACQUIRE,
> + /*weak=*/0, __ATOMIC_RELAXED,
> __ATOMIC_RELAXED);
> } while (unlikely(success == 0));
> return n;
> @@ -128,18 +134,23 @@ __rte_ring_move_cons_head(struct rte_ring *r, int
> is_sc,
> int success;
>
> /* move cons.head atomically */
> + *old_head = __atomic_load_n(&r->cons.head, __ATOMIC_RELAXED);
> do {
> /* Restore n as it may change every loop */
> n = max;
> - *old_head = __atomic_load_n(&r->cons.head,
> - __ATOMIC_ACQUIRE);
> +
> + /* this load-acquire synchronize with store-release of ht->tail
> + * in update_tail.
> + */
> + const uint32_t prod_tail = __atomic_load_n(&r->prod.tail,
> + __ATOMIC_ACQUIRE);
>
> /* The subtraction is done between two unsigned 32bits value
> * (the result is always modulo 32 bits even if we have
> * cons_head > prod_tail). So 'entries' is always between 0
> * and size(ring)-1.
> */
> - *entries = (r->prod.tail - *old_head);
> + *entries = (prod_tail - *old_head);
>
> /* Set the actual entries for dequeue */
> if (n > *entries)
> @@ -152,10 +163,11 @@ __rte_ring_move_cons_head(struct rte_ring *r, int
> is_sc,
> if (is_sc)
> r->cons.head = *new_head, success = 1;
> else
> + /* on failure, *old_head will be updated */
> success = __atomic_compare_exchange_n(&r->cons.head,
> - old_head, *new_head,
> - 0, __ATOMIC_ACQUIRE,
> - __ATOMIC_RELAXED);
> + old_head, *new_head,
> + /*weak=*/0, __ATOMIC_RELAXED,
> + __ATOMIC_RELAXED);
> } while (unlikely(success == 0));
> return n;
> }
> --
> 2.11.0
next prev parent reply other threads:[~2018-08-07 5:56 UTC|newest]
Thread overview: 131+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-06 1:18 [dpdk-dev] [PATCH] " Gavin Hu
2018-08-06 9:19 ` Thomas Monjalon
2018-08-08 1:39 ` Gavin Hu
2018-08-07 3:19 ` [dpdk-dev] [PATCH v2] " Gavin Hu
2018-08-07 5:56 ` He, Jia [this message]
2018-08-07 7:56 ` Gavin Hu
2018-08-08 3:07 ` Jerin Jacob
2018-08-08 7:23 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2018-09-17 7:47 ` [dpdk-dev] [PATCH v3 1/3] app/testpmd: show errno along with flow API errors Gavin Hu
2018-09-17 7:47 ` [dpdk-dev] [PATCH v3 2/3] net/i40e: remove invalid comment Gavin Hu
2018-09-17 8:25 ` Gavin Hu (Arm Technology China)
2018-09-17 7:47 ` [dpdk-dev] [PATCH v3 3/3] doc: add cross compile part for sample applications Gavin Hu
2018-09-17 9:48 ` Jerin Jacob
2018-09-17 10:28 ` Gavin Hu (Arm Technology China)
2018-09-17 10:34 ` Jerin Jacob
2018-09-17 10:55 ` Gavin Hu (Arm Technology China)
2018-09-17 10:49 ` [dpdk-dev] [PATCH v4] " Gavin Hu
2018-09-17 10:53 ` [dpdk-dev] [PATCH v5] " Gavin Hu
2018-09-18 11:00 ` Jerin Jacob
2018-09-19 0:33 ` [dpdk-dev] [PATCH v6] " Gavin Hu
2018-09-17 8:11 ` [dpdk-dev] [PATCH v4 1/4] bus/fslmc: fix undefined reference of memsegs Gavin Hu
2018-09-17 8:11 ` [dpdk-dev] [PATCH v4 2/4] ring: read tail using atomic load Gavin Hu
2018-09-20 6:41 ` Jerin Jacob
2018-09-25 9:26 ` Gavin Hu (Arm Technology China)
2018-09-17 8:11 ` [dpdk-dev] [PATCH v4 3/4] ring: synchronize the load and store of the tail Gavin Hu
2018-09-17 8:11 ` [dpdk-dev] [PATCH v4 4/4] ring: move the atomic load of head above the loop Gavin Hu
2018-10-27 14:21 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2018-09-17 8:17 ` [dpdk-dev] [PATCH v3 1/3] ring: read tail using atomic load Gavin Hu
2018-09-17 8:17 ` [dpdk-dev] [PATCH v3 2/3] ring: synchronize the load and store of the tail Gavin Hu
2018-09-26 9:29 ` Gavin Hu (Arm Technology China)
2018-09-26 9:59 ` Justin He
2018-09-29 10:57 ` Jerin Jacob
2018-10-17 6:29 ` [dpdk-dev] [PATCH 1/2] " Gavin Hu
2018-10-17 6:29 ` [dpdk-dev] [PATCH 2/2] ring: move the atomic load of head above the loop Gavin Hu
2018-10-17 6:35 ` [dpdk-dev] [PATCH 1/2] ring: synchronize the load and store of the tail Gavin Hu (Arm Technology China)
2018-10-27 14:39 ` Thomas Monjalon
2018-10-27 15:00 ` Jerin Jacob
2018-10-27 15:13 ` Thomas Monjalon
2018-10-27 15:34 ` Jerin Jacob
2018-10-27 15:48 ` Thomas Monjalon
2018-10-29 2:51 ` Gavin Hu (Arm Technology China)
2018-10-29 2:57 ` Gavin Hu (Arm Technology China)
2018-10-29 10:16 ` Jerin Jacob
2018-10-29 10:47 ` Thomas Monjalon
2018-10-29 11:10 ` Jerin Jacob
2018-11-03 20:12 ` Mattias Rönnblom
2018-11-05 21:51 ` Honnappa Nagarahalli
2018-11-06 11:03 ` Mattias Rönnblom
2018-10-31 3:35 ` [dpdk-dev] [PATCH v2 0/2] rte ring c11 bug fix and optimization Gavin Hu
2018-10-31 10:26 ` [dpdk-dev] [PATCH v3 0/2] ring library with c11 memory model " Gavin Hu
2018-10-31 16:58 ` Thomas Monjalon
2018-11-01 9:53 ` [dpdk-dev] [PATCH v4 1/2] ring: synchronize the load and store of the tail Gavin Hu
2018-11-01 9:53 ` [dpdk-dev] [PATCH v4 2/2] ring: move the atomic load of head above the loop Gavin Hu
2018-11-01 17:26 ` Stephen Hemminger
2018-11-02 0:53 ` Gavin Hu (Arm Technology China)
2018-11-02 4:30 ` Honnappa Nagarahalli
2018-11-02 7:15 ` Gavin Hu (Arm Technology China)
2018-11-02 9:36 ` Thomas Monjalon
2018-11-02 11:23 ` Gavin Hu (Arm Technology China)
2018-10-31 10:26 ` [dpdk-dev] [PATCH v3 1/2] ring: synchronize the load and store of the tail Gavin Hu
2018-10-31 22:07 ` Stephen Hemminger
2018-11-01 9:56 ` Gavin Hu (Arm Technology China)
2018-10-31 10:26 ` [dpdk-dev] [PATCH v3 2/2] ring: move the atomic load of head above the loop Gavin Hu
2018-10-31 3:35 ` [dpdk-dev] [PATCH v2 1/2] ring: synchronize the load and store of the tail Gavin Hu
2018-10-31 3:35 ` [dpdk-dev] [PATCH v2 2/2] ring: move the atomic load of head above the loop Gavin Hu
2018-10-31 9:36 ` Thomas Monjalon
2018-10-31 10:27 ` Gavin Hu (Arm Technology China)
2018-11-01 9:53 ` [dpdk-dev] [PATCH v4 0/2] ring library with c11 memory model bug fix and optimization Gavin Hu
2018-11-02 11:21 ` [dpdk-dev] [PATCH v5 " Gavin Hu
2018-11-02 11:21 ` [dpdk-dev] [PATCH v5 1/2] ring: synchronize the load and store of the tail Gavin Hu
2018-11-05 9:30 ` Olivier Matz
2018-11-02 11:21 ` [dpdk-dev] [PATCH v5 2/2] ring: move the atomic load of head above the loop Gavin Hu
2018-11-02 11:43 ` Bruce Richardson
2018-11-03 1:19 ` Gavin Hu (Arm Technology China)
2018-11-03 9:34 ` Honnappa Nagarahalli
2018-11-05 13:17 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2018-11-05 13:41 ` Jerin Jacob
2018-11-05 9:44 ` [dpdk-dev] " Olivier Matz
2018-11-05 13:36 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
2018-09-17 8:17 ` [dpdk-dev] [PATCH v3 3/3] " Gavin Hu
2018-09-26 9:29 ` Gavin Hu (Arm Technology China)
2018-09-26 10:06 ` Justin He
2018-09-29 7:19 ` Stephen Hemminger
2018-09-29 10:59 ` Jerin Jacob
2018-09-26 9:29 ` [dpdk-dev] [PATCH v3 1/3] ring: read tail using atomic load Gavin Hu (Arm Technology China)
2018-09-26 10:09 ` Justin He
2018-09-29 10:48 ` Jerin Jacob
2018-10-05 0:47 ` Gavin Hu (Arm Technology China)
2018-10-05 8:21 ` Ananyev, Konstantin
2018-10-05 11:15 ` Ola Liljedahl
2018-10-05 11:36 ` Ola Liljedahl
2018-10-05 13:44 ` Ananyev, Konstantin
2018-10-05 14:21 ` Ola Liljedahl
2018-10-05 15:11 ` Honnappa Nagarahalli
2018-10-05 17:07 ` Jerin Jacob
2018-10-05 18:05 ` Ola Liljedahl
2018-10-05 20:06 ` Honnappa Nagarahalli
2018-10-05 20:17 ` Ola Liljedahl
2018-10-05 20:29 ` Honnappa Nagarahalli
2018-10-05 20:34 ` Ola Liljedahl
2018-10-06 7:41 ` Jerin Jacob
2018-10-06 19:44 ` Ola Liljedahl
2018-10-06 19:59 ` Ola Liljedahl
2018-10-07 4:02 ` Jerin Jacob
2018-10-07 20:11 ` Ola Liljedahl
2018-10-07 20:44 ` Ola Liljedahl
2018-10-08 6:06 ` Jerin Jacob
2018-10-08 9:22 ` Ola Liljedahl
2018-10-08 10:00 ` Jerin Jacob
2018-10-08 10:25 ` Ola Liljedahl
2018-10-08 10:33 ` Gavin Hu (Arm Technology China)
2018-10-08 10:39 ` Ola Liljedahl
2018-10-08 10:41 ` Gavin Hu (Arm Technology China)
2018-10-08 10:49 ` Jerin Jacob
2018-10-10 6:28 ` Gavin Hu (Arm Technology China)
2018-10-10 19:26 ` Honnappa Nagarahalli
2018-10-08 10:46 ` Jerin Jacob
2018-10-08 11:21 ` Ola Liljedahl
2018-10-08 11:50 ` Jerin Jacob
2018-10-08 11:59 ` Ola Liljedahl
2018-10-08 12:05 ` Jerin Jacob
2018-10-08 12:20 ` Jerin Jacob
2018-10-08 12:30 ` Ola Liljedahl
2018-10-09 8:53 ` Olivier Matz
2018-10-09 3:16 ` Honnappa Nagarahalli
2018-10-08 14:43 ` Bruce Richardson
2018-10-08 14:46 ` Ola Liljedahl
2018-10-08 15:45 ` Ola Liljedahl
2018-10-08 5:27 ` Honnappa Nagarahalli
2018-10-08 10:01 ` Ola Liljedahl
2018-10-27 14:17 ` [dpdk-dev] [dpdk-stable] " Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=88b4421b63ec4802892f3fc40d457f03@HXTBJIDCEMVIW01.hxtcorp.net \
--to=jia.he@hxt-semitech.com \
--cc=Honnappa.Nagarahalli@arm.com \
--cc=Ola.Liljedahl@arm.com \
--cc=dev@dpdk.org \
--cc=gavin.hu@arm.com \
--cc=hemant.agrawal@nxp.com \
--cc=jerin.jacob@caviumnetworks.com \
--cc=stable@dpdk.org \
--cc=steve.capper@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).