From: Phil Yang <Phil.Yang@arm.com>
To: "pbhagavatula@marvell.com" <pbhagavatula@marvell.com>,
"jerinj@marvell.com" <jerinj@marvell.com>,
Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>, nd <nd@arm.com>, nd <nd@arm.com>
Subject: Re: [dpdk-dev] [PATCH] event/octeontx2: use c11 atomics for statistics
Date: Mon, 30 Mar 2020 06:25:42 +0000 [thread overview]
Message-ID: <VE1PR08MB464085320F9D499D4F1F6060E9CB0@VE1PR08MB4640.eurprd08.prod.outlook.com> (raw)
In-Reply-To: <20200327085558.1136-1-pbhagavatula@marvell.com>
> -----Original Message-----
> From: pbhagavatula@marvell.com <pbhagavatula@marvell.com>
> Sent: Friday, March 27, 2020 4:56 PM
> To: jerinj@marvell.com; Honnappa Nagarahalli
> <Honnappa.Nagarahalli@arm.com>; Phil Yang <Phil.Yang@arm.com>; Pavan
> Nikhilesh <pbhagavatula@marvell.com>
> Cc: dev@dpdk.org
> Subject: [dpdk-dev] [PATCH] event/octeontx2: use c11 atomics for statistics
>
> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
> Use c11 atomics with RELAXED ordering instead of rte_atomic ops which
> enforce unnessary barries on arm64.
>
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
> ---
> drivers/event/octeontx2/otx2_tim_evdev.c | 5 +++--
> drivers/event/octeontx2/otx2_tim_evdev.h | 2 +-
> drivers/event/octeontx2/otx2_tim_worker.c | 5 +++--
> 3 files changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/event/octeontx2/otx2_tim_evdev.c
> b/drivers/event/octeontx2/otx2_tim_evdev.c
> index cd0dcde24..4c24cc8a6 100644
> --- a/drivers/event/octeontx2/otx2_tim_evdev.c
> +++ b/drivers/event/octeontx2/otx2_tim_evdev.c
> @@ -526,7 +526,8 @@ otx2_tim_stats_get(const struct
> rte_event_timer_adapter *adapter,
> uint64_t bkt_cyc = rte_rdtsc() - tim_ring->ring_start_cyc;
>
>
> - stats->evtim_exp_count = rte_atomic64_read(&tim_ring->arm_cnt);
> + stats->evtim_exp_count = __atomic_load_n(&tim_ring->arm_cnt,
> + __ATOMIC_RELAXED);
> stats->ev_enq_count = stats->evtim_exp_count;
> stats->adapter_tick_count = rte_reciprocal_divide_u64(bkt_cyc,
> &tim_ring->fast_div);
> @@ -538,7 +539,7 @@ otx2_tim_stats_reset(const struct
> rte_event_timer_adapter *adapter)
> {
> struct otx2_tim_ring *tim_ring = adapter->data->adapter_priv;
>
> - rte_atomic64_clear(&tim_ring->arm_cnt);
> + __atomic_store_n(&tim_ring->arm_cnt, 0, __ATOMIC_RELAXED);
Both otx2_tim_stats_get & otx2_tim_stats_reset operations are handled in the same thread,
and the arm_cmn read & store operations are sequential consistent in this case.
So RELAXED memory ordering here is enough.
> return 0;
> }
>
> diff --git a/drivers/event/octeontx2/otx2_tim_evdev.h
> b/drivers/event/octeontx2/otx2_tim_evdev.h
> index 56895dcbf..44e3c7b51 100644
> --- a/drivers/event/octeontx2/otx2_tim_evdev.h
> +++ b/drivers/event/octeontx2/otx2_tim_evdev.h
> @@ -149,7 +149,7 @@ struct otx2_tim_ring {
> struct otx2_tim_bkt *bkt;
> struct rte_mempool *chunk_pool;
> struct rte_reciprocal_u64 fast_div;
> - rte_atomic64_t arm_cnt;
> + uint64_t arm_cnt;
> uint8_t prod_type_sp;
> uint8_t enable_stats;
> uint8_t disable_npa;
> diff --git a/drivers/event/octeontx2/otx2_tim_worker.c
> b/drivers/event/octeontx2/otx2_tim_worker.c
> index 104674c79..4b5cfdc72 100644
> --- a/drivers/event/octeontx2/otx2_tim_worker.c
> +++ b/drivers/event/octeontx2/otx2_tim_worker.c
> @@ -88,7 +88,7 @@ tim_timer_arm_burst(const struct
> rte_event_timer_adapter *adptr,
> }
>
> if (flags & OTX2_TIM_ENA_STATS)
> - rte_atomic64_add(&tim_ring->arm_cnt, index);
> + __atomic_fetch_add(&tim_ring->arm_cnt, index,
> __ATOMIC_RELAXED);
>
> return index;
> }
> @@ -130,7 +130,8 @@ tim_timer_arm_tmo_brst(const struct
> rte_event_timer_adapter *adptr,
> break;
> }
> if (flags & OTX2_TIM_ENA_STATS)
> - rte_atomic64_add(&tim_ring->arm_cnt, set_timers);
> + __atomic_fetch_add(&tim_ring->arm_cnt, set_timers,
> + __ATOMIC_RELAXED);
>
> return set_timers;
> }
> --
> 2.17.1
It goods good to me.
Reviewed-by: Phil Yang <phil.yang@arm.com>
Thanks,
Phil
next prev parent reply other threads:[~2020-03-30 6:25 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-03-27 8:55 pbhagavatula
2020-03-30 6:25 ` Phil Yang [this message]
2020-04-04 16:39 ` Jerin Jacob
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=VE1PR08MB464085320F9D499D4F1F6060E9CB0@VE1PR08MB4640.eurprd08.prod.outlook.com \
--to=phil.yang@arm.com \
--cc=Honnappa.Nagarahalli@arm.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=nd@arm.com \
--cc=pbhagavatula@marvell.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).