DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Van Haaren, Harry" <harry.van.haaren@intel.com>
To: "dev@dpdk.org" <dev@dpdk.org>
Cc: "david.marchand@redhat.com" <david.marchand@redhat.com>,
	mattias.ronnblom <mattias.ronnblom@ericsson.com>,
	"jerinj@marvell.com" <jerinj@marvell.com>
Subject: Re: [dpdk-dev] [PATCH v4] event/sw: add xstats to expose progress details
Date: Thu, 4 Mar 2021 11:02:33 +0000	[thread overview]
Message-ID: <BYAPR11MB314337E51B48EB4A5E43E664D7979@BYAPR11MB3143.namprd11.prod.outlook.com> (raw)
In-Reply-To: <20210303105643.2552378-1-harry.van.haaren@intel.com>

Fix typo in Mattias' email, apologies for noise.

> -----Original Message-----
> From: Van Haaren, Harry <harry.van.haaren@intel.com>
> Sent: Wednesday, March 3, 2021 10:57 AM
> To: dev@dpdk.org
> Cc: david.marchand@redhat.com; mattias.ronnblom@ericcson.com;
> jerinj@marvell.com; Van Haaren, Harry <harry.van.haaren@intel.com>
> Subject: [PATCH v4] event/sw: add xstats to expose progress details
> 
> Today it is difficult to know if the SW Eventdev PMD is making
> forward progress when it runs an iteration of its service. This
> commit adds two xstats to give better visibility to the application.
> 
> The new xstats provide an application with which Eventdev ports
> recieved work in the last iteration of scheduling, as well if
> forward progress was made by the scheduler.
> 
> This patch implements an xstat for the SW PMD that exposes a
> bitmask of ports that were scheduled to. In the unlikely case
> that the SW PMD instance has 64 or more ports, return UINT64_MAX.
> 
> Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
> 
> ---
> 
> v3:
> - Simplify all metrics to Event SW PMD
> 
> v2:
> - Fixup printf() %ld to PRIu64
> 
> Note most of the changes here are unit-test changes to add
> a statistic to the PMD. The actual "useful code" is a mere
> handful of lines in a lot of noise.
> 
> ---
>  drivers/event/sw/sw_evdev.h           |  2 ++
>  drivers/event/sw/sw_evdev_scheduler.c | 15 ++++++++++++++
>  drivers/event/sw/sw_evdev_selftest.c  | 28 ++++++++++++++-------------
>  drivers/event/sw/sw_evdev_xstats.c    |  9 ++++++++-
>  4 files changed, 40 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/event/sw/sw_evdev.h b/drivers/event/sw/sw_evdev.h
> index 5ab6465c83..33645bd1df 100644
> --- a/drivers/event/sw/sw_evdev.h
> +++ b/drivers/event/sw/sw_evdev.h
> @@ -259,6 +259,8 @@ struct sw_evdev {
>  	uint64_t sched_no_iq_enqueues;
>  	uint64_t sched_no_cq_enqueues;
>  	uint64_t sched_cq_qid_called;
> +	uint64_t sched_last_iter_bitmask;
> +	uint8_t sched_progress_last_iter;
> 
>  	uint8_t started;
>  	uint32_t credit_update_quanta;
> diff --git a/drivers/event/sw/sw_evdev_scheduler.c
> b/drivers/event/sw/sw_evdev_scheduler.c
> index f747b3c6d4..d3a6bd5cda 100644
> --- a/drivers/event/sw/sw_evdev_scheduler.c
> +++ b/drivers/event/sw/sw_evdev_scheduler.c
> @@ -559,6 +559,11 @@ sw_event_schedule(struct rte_eventdev *dev)
>  	sw->sched_no_iq_enqueues += (in_pkts_total == 0);
>  	sw->sched_no_cq_enqueues += (out_pkts_total == 0);
> 
> +	uint64_t work_done = (in_pkts_total + out_pkts_total) != 0;
> +	sw->sched_progress_last_iter = work_done;
> +
> +	uint64_t cqs_scheds_last_iter = 0;
> +
>  	/* push all the internal buffered QEs in port->cq_ring to the
>  	 * worker cores: aka, do the ring transfers batched.
>  	 */
> @@ -578,6 +583,7 @@ sw_event_schedule(struct rte_eventdev *dev)
>  					&sw->cq_ring_space[i]);
>  			port->cq_buf_count = 0;
>  			no_enq = 0;
> +			cqs_scheds_last_iter |= (1ULL << i);
>  		} else {
>  			sw->cq_ring_space[i] =
>  					rte_event_ring_free_count(worker) -
> @@ -597,4 +603,13 @@ sw_event_schedule(struct rte_eventdev *dev)
>  			sw->sched_min_burst = sw->sched_min_burst_size;
>  	}
> 
> +	/* Provide stats on what eventdev ports were scheduled to this
> +	 * iteration. If more than 64 ports are active, always report that
> +	 * all Eventdev ports have been scheduled events.
> +	 */
> +	if (likely(sw->port_count < 64)) {
> +		sw->sched_last_iter_bitmask = cqs_scheds_last_iter;
> +	} else {
> +		sw->sched_last_iter_bitmask = UINT64_MAX;
> +	}
>  }
> diff --git a/drivers/event/sw/sw_evdev_selftest.c
> b/drivers/event/sw/sw_evdev_selftest.c
> index e4bfb3a0f1..d53e903129 100644
> --- a/drivers/event/sw/sw_evdev_selftest.c
> +++ b/drivers/event/sw/sw_evdev_selftest.c
> @@ -873,15 +873,15 @@ xstats_tests(struct test *t)
>  	int ret = rte_event_dev_xstats_names_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_DEVICE,
>  					0, xstats_names, ids, XSTATS_MAX);
> -	if (ret != 6) {
> -		printf("%d: expected 6 stats, got return %d\n", __LINE__, ret);
> +	if (ret != 8) {
> +		printf("%d: expected 8 stats, got return %d\n", __LINE__, ret);
>  		return -1;
>  	}
>  	ret = rte_event_dev_xstats_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_DEVICE,
>  					0, ids, values, ret);
> -	if (ret != 6) {
> -		printf("%d: expected 6 stats, got return %d\n", __LINE__, ret);
> +	if (ret != 8) {
> +		printf("%d: expected 8 stats, got return %d\n", __LINE__, ret);
>  		return -1;
>  	}
> 
> @@ -959,7 +959,7 @@ xstats_tests(struct test *t)
>  	ret = rte_event_dev_xstats_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_DEVICE,
>  					0, ids, values, num_stats);
> -	static const uint64_t expected[] = {3, 3, 0, 1, 0, 0};
> +	static const uint64_t expected[] = {3, 3, 0, 1, 0, 0, 4, 1};
>  	for (i = 0; (signed int)i < ret; i++) {
>  		if (expected[i] != values[i]) {
>  			printf(
> @@ -975,7 +975,7 @@ xstats_tests(struct test *t)
>  					0, NULL, 0);
> 
>  	/* ensure reset statistics are zero-ed */
> -	static const uint64_t expected_zero[] = {0, 0, 0, 0, 0, 0};
> +	static const uint64_t expected_zero[] = {0, 0, 0, 0, 0, 0, 0, 0};
>  	ret = rte_event_dev_xstats_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_DEVICE,
>  					0, ids, values, num_stats);
> @@ -1460,7 +1460,7 @@ xstats_id_reset_tests(struct test *t)
>  	for (i = 0; i < XSTATS_MAX; i++)
>  		ids[i] = i;
> 
> -#define NUM_DEV_STATS 6
> +#define NUM_DEV_STATS 8
>  	/* Device names / values */
>  	int num_stats = rte_event_dev_xstats_names_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_DEVICE,
> @@ -1504,8 +1504,10 @@ xstats_id_reset_tests(struct test *t)
>  	static const char * const dev_names[] = {
>  		"dev_rx", "dev_tx", "dev_drop", "dev_sched_calls",
>  		"dev_sched_no_iq_enq", "dev_sched_no_cq_enq",
> +		"dev_sched_last_iter_bitmask",
> +		"dev_sched_progress_last_iter"
>  	};
> -	uint64_t dev_expected[] = {NPKTS, NPKTS, 0, 1, 0, 0};
> +	uint64_t dev_expected[] = {NPKTS, NPKTS, 0, 1, 0, 0, 4, 1};
>  	for (i = 0; (int)i < ret; i++) {
>  		unsigned int id;
>  		uint64_t val = rte_event_dev_xstats_by_name_get(evdev,
> @@ -1518,8 +1520,8 @@ xstats_id_reset_tests(struct test *t)
>  		}
>  		if (val != dev_expected[i]) {
>  			printf("%d: %s value incorrect, expected %"
> -				PRIu64" got %d\n", __LINE__, dev_names[i],
> -				dev_expected[i], id);
> +				PRIu64" got %"PRIu64"\n", __LINE__,
> +				dev_names[i], dev_expected[i], val);
>  			goto fail;
>  		}
>  		/* reset to zero */
> @@ -1542,11 +1544,11 @@ xstats_id_reset_tests(struct test *t)
>  		}
>  	};
> 
> -/* 48 is stat offset from start of the devices whole xstats.
> +/* 49 is stat offset from start of the devices whole xstats.
>   * This WILL break every time we add a statistic to a port
>   * or the device, but there is no other way to test
>   */
> -#define PORT_OFF 48
> +#define PORT_OFF 50
>  /* num stats for the tested port. CQ size adds more stats to a port */
>  #define NUM_PORT_STATS 21
>  /* the port to test. */
> @@ -1670,7 +1672,7 @@ xstats_id_reset_tests(struct test *t)
>  /* queue offset from start of the devices whole xstats.
>   * This will break every time we add a statistic to a device/port/queue
>   */
> -#define QUEUE_OFF 90
> +#define QUEUE_OFF 92
>  	const uint32_t queue = 0;
>  	num_stats = rte_event_dev_xstats_names_get(evdev,
>  					RTE_EVENT_DEV_XSTATS_QUEUE, queue,
> diff --git a/drivers/event/sw/sw_evdev_xstats.c
> b/drivers/event/sw/sw_evdev_xstats.c
> index 02f7874180..c2647d7da2 100644
> --- a/drivers/event/sw/sw_evdev_xstats.c
> +++ b/drivers/event/sw/sw_evdev_xstats.c
> @@ -17,6 +17,8 @@ enum xstats_type {
>  	/* device instance specific */
>  	no_iq_enq,
>  	no_cq_enq,
> +	sched_last_iter_bitmask,
> +	sched_progress_last_iter,
>  	/* port_specific */
>  	rx_used,
>  	rx_free,
> @@ -57,6 +59,9 @@ get_dev_stat(const struct sw_evdev *sw, uint16_t obj_idx
> __rte_unused,
>  	case calls: return sw->sched_called;
>  	case no_iq_enq: return sw->sched_no_iq_enqueues;
>  	case no_cq_enq: return sw->sched_no_cq_enqueues;
> +	case sched_last_iter_bitmask: return sw->sched_last_iter_bitmask;
> +	case sched_progress_last_iter: return sw->sched_progress_last_iter;
> +
>  	default: return -1;
>  	}
>  }
> @@ -177,9 +182,11 @@ sw_xstats_init(struct sw_evdev *sw)
>  	 */
>  	static const char * const dev_stats[] = { "rx", "tx", "drop",
>  			"sched_calls", "sched_no_iq_enq", "sched_no_cq_enq",
> +			"sched_last_iter_bitmask", "sched_progress_last_iter",
>  	};
>  	static const enum xstats_type dev_types[] = { rx, tx, dropped,
> -			calls, no_iq_enq, no_cq_enq,
> +			calls, no_iq_enq, no_cq_enq, sched_last_iter_bitmask,
> +			sched_progress_last_iter,
>  	};
>  	/* all device stats are allowed to be reset */
> 
> --
> 2.25.1


  reply	other threads:[~2021-03-04 11:02 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-12 16:58 [dpdk-dev] [PATCH v3 1/3] service: add component useful work attribute Harry van Haaren
2021-02-12 16:58 ` [dpdk-dev] [PATCH v3 2/3] event/sw: add useful work done attribute Harry van Haaren
2021-02-12 16:58 ` [dpdk-dev] [PATCH v3 3/3] event/sw: add xstat for work done in last iteration Harry van Haaren
2021-03-03 10:56 ` [dpdk-dev] [PATCH v4] event/sw: add xstats to expose progress details Harry van Haaren
2021-03-04 11:02   ` Van Haaren, Harry [this message]
2021-03-08  7:52     ` Jerin Jacob
2021-03-08  9:22   ` [dpdk-dev] [PATCH v5] " Harry van Haaren
2021-03-08 16:54     ` 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=BYAPR11MB314337E51B48EB4A5E43E664D7979@BYAPR11MB3143.namprd11.prod.outlook.com \
    --to=harry.van.haaren@intel.com \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    --cc=jerinj@marvell.com \
    --cc=mattias.ronnblom@ericsson.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).