From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>
Subject: Re: [PATCH] ether: add support for vtune task tracing
Date: Thu, 22 Jun 2017 09:42:20 +0000
Message-ID: <2601191342CEEE43887BDE71AB9772583FB0BD8E@IRSMSX109.ger.corp.intel.com>
References: <1497892689-27494-1-git-send-email-ilia.kurakin@intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Cc: "Kurakin, Ilia" <ilia.kurakin@intel.com>
To: "Kurakin, Ilia" <ilia.kurakin@intel.com>, "dev@dpdk.org" <dev@dpdk.org>
Return-path: <dev-bounces@dpdk.org>
Received: from mga01.intel.com (mga01.intel.com [192.55.52.88])
 by dpdk.org (Postfix) with ESMTP id 784E05688
 for <dev@dpdk.org>; Thu, 22 Jun 2017 11:42:41 +0200 (CEST)
In-Reply-To: <1497892689-27494-1-git-send-email-ilia.kurakin@intel.com>
Content-Language: en-US
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

Hi Ilia,

> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of ilia.kurakin@intel.c=
om
> Sent: Monday, June 19, 2017 6:18 PM
> To: dev@dpdk.org
> Cc: Kurakin, Ilia <ilia.kurakin@intel.com>
> Subject: [dpdk-dev] [PATCH] ether: add support for vtune task tracing
>=20
> From: Ilia Kurakin <ilia.kurakin@intel.com>
>=20
> The patch adds tracing of loop iterations that yielded no packets in a DP=
DK
> application. It is using ITT task API:
>     https://software.intel.com/en-us/node/544206
>=20
> We suppose the flow of using this tracing would assume the user has ITT l=
ib
> and header on his machine and re-build DPDK with additional make paramete=
rs:
>=20
>     make EXTRA_CFLAGS=3D-I<path to ittnotify.h>
>          EXTRA_LDLIBS=3D"-L<path to libittnotify.a> -littnotify"

There are few things that worry me with that patch:
1. We add new config variable and add extra dependency here.
    Usually we try not to do that without really compelling reason.
2. We pollute rte_ethdev with the code that has nothing to do with
     it major functionality.

That makes me wonder why this vtune data collection has to be done
inside rx_burst() function?
 Why it can't be done on the application layer, i.e. straight after rx_burs=
t()
is finished?
Something like:
n =3D rte_eth_rx_burst(port, queue, ....);
itt_rx_collect_data(port, queue, n, ....);
?

Or as alternative, user can install this vtune collection routine as
rx callback function.

Konstantin

>=20
> Signed-off-by: Ilia Kurakin <ilia.kurakin@intel.com>
> ---
>  config/common_base             |  1 +
>  lib/librte_ether/Makefile      |  1 +
>  lib/librte_ether/rte_eth_itt.h | 69 ++++++++++++++++++++++++++++++++++++=
++++++
>  lib/librte_ether/rte_ethdev.c  |  7 +++++
>  lib/librte_ether/rte_ethdev.h  | 26 ++++++++++++++++
>  5 files changed, 104 insertions(+)
>  create mode 100644 lib/librte_ether/rte_eth_itt.h
>=20
> diff --git a/config/common_base b/config/common_base
> index f6aafd1..60d8b63 100644
> --- a/config/common_base
> +++ b/config/common_base
> @@ -135,6 +135,7 @@ CONFIG_RTE_MAX_QUEUES_PER_PORT=3D1024
>  CONFIG_RTE_LIBRTE_IEEE1588=3Dn
>  CONFIG_RTE_ETHDEV_QUEUE_STAT_CNTRS=3D16
>  CONFIG_RTE_ETHDEV_RXTX_CALLBACKS=3Dy
> +CONFIG_RTE_ETHDEV_TRACE_WASTED_RX_ITERATIONS=3Dn
>=20
>  #
>  # Turn off Tx preparation stage
> diff --git a/lib/librte_ether/Makefile b/lib/librte_ether/Makefile
> index 93fdde1..c10153a 100644
> --- a/lib/librte_ether/Makefile
> +++ b/lib/librte_ether/Makefile
> @@ -56,5 +56,6 @@ SYMLINK-y-include +=3D rte_eth_ctrl.h
>  SYMLINK-y-include +=3D rte_dev_info.h
>  SYMLINK-y-include +=3D rte_flow.h
>  SYMLINK-y-include +=3D rte_flow_driver.h
> +SYMLINK-${CONFIG_RTE_ETHDEV_TRACE_WASTED_RX_ITERATIONS}-include +=3D rte=
_eth_itt.h
>=20
>  include $(RTE_SDK)/mk/rte.lib.mk
> diff --git a/lib/librte_ether/rte_eth_itt.h b/lib/librte_ether/rte_eth_it=
t.h
> new file mode 100644
> index 0000000..e7984fb
> --- /dev/null
> +++ b/lib/librte_ether/rte_eth_itt.h
> @@ -0,0 +1,69 @@
> +#ifndef _RTE_ETH_ITT_H_
> +#define _RTE_ETH_ITT_H_
> +
> +#include <ittnotify.h>
> +#include <rte_config.h>
> +
> +#define ITT_MAX_NAME_LEN (100)
> +
> +/**
> + * Auxiliary ITT structure belonging to port and using to:
> + *   -  track queue state to determine whether it is wasting loop iterat=
ions
> + *   -  begin or end ITT task using task domain and name
> + */
> +struct rte_eth_itt_aux_data {
> +	/**
> +	 * ITT domains for each queue.
> +	 */
> +	__itt_domain *wasted_iteration_itt_domains[RTE_MAX_QUEUES_PER_PORT];
> +	/**
> +	 * ITT task names for each queue.
> +	 */
> +	__itt_string_handle *wasted_iteration_itt_handles[RTE_MAX_QUEUES_PER_PO=
RT];
> +	/**
> +	 * Flags indicating the queues state. Possible values:
> +	 * 1 - queue is wasting iterations, 0 - otherwise.
> +	 */
> +	uint8_t queue_is_wasting_iterations[RTE_MAX_QUEUES_PER_PORT];
> +};
> +
> +/**
> + * The pool of *rte_eth_itt_aux_data* structures.
> + */
> +struct rte_eth_itt_aux_data itt_aux_data[RTE_MAX_ETHPORTS];
> +
> +/**
> + * Initialization of rte_eth_itt_aux_data for a given port.
> + * This function must be invoked when ethernet device is being configure=
d.
> + * Result will be stored in the global array *itt_aux_data*.
> + *
> + * @param port_id
> + *  The port identifier of the Ethernet device.
> + * @param port_name
> + *  The name of the Ethernet device.
> + * @param queue_num
> + *  The number of queues on specified port.
> + */
> +static inline void
> +rte_eth_init_itt(uint8_t port_id, char *port_name, uint8_t queue_num) {
> +	uint16_t q_id;
> +	for (q_id =3D 0; q_id < queue_num; ++q_id) {
> +		char domain_name[ITT_MAX_NAME_LEN];
> +		snprintf(domain_name, sizeof(domain_name),
> +			"RXBurst.WastedIterations.Port_%s.Queue_%d",
> +			port_name, q_id);
> +		itt_aux_data[port_id].wasted_iteration_itt_domains[q_id]
> +			=3D __itt_domain_create(domain_name);
> +
> +		char task_name[ITT_MAX_NAME_LEN];
> +		snprintf(task_name, sizeof(task_name),
> +			"port id: %d; queue id: %d",
> +			port_id, q_id);
> +		itt_aux_data[port_id].wasted_iteration_itt_handles[q_id]
> +			=3D __itt_string_handle_create(task_name);
> +
> +		itt_aux_data[port_id].queue_is_wasting_iterations[q_id] =3D 0;
> +	}
> +}
> +
> +#endif
> diff --git a/lib/librte_ether/rte_ethdev.c b/lib/librte_ether/rte_ethdev.=
c
> index 81a45c0..9e5ac01 100644
> --- a/lib/librte_ether/rte_ethdev.c
> +++ b/lib/librte_ether/rte_ethdev.c
> @@ -818,6 +818,13 @@ rte_eth_dev_configure(uint8_t port_id, uint16_t nb_r=
x_q, uint16_t nb_tx_q,
>  		return diag;
>  	}
>=20
> +#ifdef RTE_ETHDEV_TRACE_WASTED_RX_ITERATIONS
> +	/**
> +	* See rte_eth_itt.h to find comments on code below.
> +	*/
> +	rte_eth_init_itt(port_id, dev->data->name, nb_rx_q);
> +#endif
> +
>  	return 0;
>  }
>=20
> diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.=
h
> index f6e6c74..4ba90d2 100644
> --- a/lib/librte_ether/rte_ethdev.h
> +++ b/lib/librte_ether/rte_ethdev.h
> @@ -186,6 +186,10 @@ extern "C" {
>  #include "rte_eth_ctrl.h"
>  #include "rte_dev_info.h"
>=20
> +#ifdef RTE_ETHDEV_TRACE_WASTED_RX_ITERATIONS
> +#include "rte_eth_itt.h"
> +#endif
> +
>  struct rte_mbuf;
>=20
>  /**
> @@ -2710,6 +2714,28 @@ rte_eth_rx_burst(uint8_t port_id, uint16_t queue_i=
d,
>  	int16_t nb_rx =3D (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
>  			rx_pkts, nb_pkts);
>=20
> +#ifdef RTE_ETHDEV_TRACE_WASTED_RX_ITERATIONS
> +	/**
> +	* See rte_eth_itt.h to find comments on code below.
> +	*/
> +	if (unlikely(nb_rx =3D=3D 0)) {
> +		if (!itt_aux_data[port_id].queue_is_wasting_iterations[queue_id]) {
> +			__itt_task_begin(
> +				itt_aux_data[port_id].wasted_iteration_itt_domains[queue_id],
> +				__itt_null, __itt_null,
> +				itt_aux_data[port_id].wasted_iteration_itt_handles[queue_id]);
> +			itt_aux_data[port_id].queue_is_wasting_iterations[queue_id] =3D 1;
> +		}
> +	}
> +	else {
> +		if (unlikely(itt_aux_data[port_id].queue_is_wasting_iterations[queue_i=
d])) {
> +			__itt_task_end(
> +				itt_aux_data[port_id].wasted_iteration_itt_domains[queue_id]);
> +			itt_aux_data[port_id].queue_is_wasting_iterations[queue_id] =3D 0;
> +		}
> +	}
> +#endif
> +
>  #ifdef RTE_ETHDEV_RXTX_CALLBACKS
>  	struct rte_eth_rxtx_callback *cb =3D dev->post_rx_burst_cbs[queue_id];
>=20
> --
> 2.7.4
>=20
>=20
> --------------------------------------------------------------------
> Joint Stock Company Intel A/O
> Registered legal address: Krylatsky Hills Business Park,
> 17 Krylatskaya Str., Bldg 4, Moscow 121614,
> Russian Federation
>=20
> This e-mail and any attachments may contain confidential material for
> the sole use of the intended recipient(s). Any review or distribution
> by others is strictly prohibited. If you are not the intended
> recipient, please contact the sender and delete all copies.