From: "Morten Brørup" <mb@smartsharesystems.com>
To: "Hideyuki Yamashita" <yamashita.hideyuki@ntt-tx.co.jp>,
"Ananyev, Konstantin" <konstantin.ananyev@intel.com>
Cc: "Thomas Monjalon" <thomas@monjalon.net>,
"Yigit, Ferruh" <ferruh.yigit@intel.com>,
"Andrew Rybchenko" <andrew.rybchenko@oktetlabs.ru>,
"Ray Kinsella" <mdr@ashroe.eu>,
"Neil Horman" <nhorman@tuxdriver.com>,
"Burakov, Anatoly" <anatoly.burakov@intel.com>, <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 5/5] librte_ethdev: add to use apistats
Date: Tue, 22 Dec 2020 10:04:19 +0100 [thread overview]
Message-ID: <98CBD80474FA8B44BF855DF32C47DC35C614ED@smartserver.smartshare.dk> (raw)
In-Reply-To: <20201222115006.CFA6.17218CA3@ntt-tx.co.jp>
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Hideyuki Yamashita
> Sent: Tuesday, December 22, 2020 3:50 AM
>
> Hello,
>
> Thanks for your comments.
> Please see my comments inline tagged with [HY].
>
> > > diff --git a/lib/librte_ethdev/rte_ethdev.h
> b/lib/librte_ethdev/rte_ethdev.h
> > > index f5f8919..bef9bc6 100644
> > > --- a/lib/librte_ethdev/rte_ethdev.h
> > > +++ b/lib/librte_ethdev/rte_ethdev.h
> > > @@ -160,6 +160,7 @@ extern "C" {
> > >
> > > #include "rte_ethdev_trace_fp.h"
> > > #include "rte_dev_info.h"
> > > +#include <rte_apistats.h>
> > >
> > > extern int rte_eth_dev_logtype;
> > >
> > > @@ -4849,6 +4850,9 @@ rte_eth_rx_burst(uint16_t port_id, uint16_t
> queue_id,
> > > nb_rx = (*dev->rx_pkt_burst)(dev->data->rx_queues[queue_id],
> > > rx_pkts, nb_pkts);
> > >
> > > + int lcore_id = rte_lcore_id();
> > > + rte_apicounts->rx_burst_counts[lcore_id]++;
[...]
> > As a generic one: such sort of statistics can be easily collected by the
> app itself.
> > Either by just incrementing counters before rx/tx_burst function call
> directly or
> > via rx/tx callbacks.
> > So I don't see any point to push it inside ethdev layer.
> > Konstantin
> [HY]
> You are correct.
> Application can do it.
> But if applications want to do it, then every applicaiton needs
> to do it.
> The reason why I propose my patchset is to provide such
> common function (count invocation of rx_burst/tx_burst)
> so that application only needs to "retrieve the information".
>
> I think it is benefitical than all application do similar thing.
> Your feedback is highly appreciated.
For performance reasons, I am strongly opposed to adding anything into the ethdev rx/tx functions, unless it is absolutely critical for a large user base.
I get your argument about avoiding additional application code by doing it directly in the ethdev rx/tx functions - this is the benefit that this library adds to DPDK. So as a compromise, I suggest surrounding the added code in these functions by #ifdef/#endif, so there is no performance degradation if the library is not used.
Personally, I would prefer a much more advanced library for measuring CPU load and RX/TX usage. However, I can also imagine that simple DPDK applications would benefit from this library, because is easy to understand and requires nearly no effort to use.
>
> Thanks!
> BR,
> Hideyuki Yamashita
> NTT TechnoCross
>
> > > #ifdef RTE_ETHDEV_RXTX_CALLBACKS
> > > struct rte_eth_rxtx_callback *cb;
> > >
> > > @@ -5124,6 +5128,9 @@ rte_eth_tx_burst(uint16_t port_id, uint16_t
> queue_id,
> > > }
> > > #endif
> > >
> > > + int lcore_id = rte_lcore_id();
> > > + rte_apicounts->tx_burst_counts[lcore_id]++;
> > > #ifdef RTE_ETHDEV_RXTX_CALLBACKS
> > > struct rte_eth_rxtx_callback *cb;
> > >
next prev parent reply other threads:[~2020-12-22 9:04 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-12-04 7:51 [dpdk-dev] [PATCH 0/5] add apistats function Hideyuki Yamashita
2020-12-04 7:51 ` [dpdk-dev] [PATCH 1/5] maintainers: update maintainers file for apistats Hideyuki Yamashita
2020-12-05 13:27 ` Varghese, Vipin
2020-12-24 6:36 ` Hideyuki Yamashita
2020-12-04 7:51 ` [dpdk-dev] [PATCH 2/5] app/proc-info: add to use apistats Hideyuki Yamashita
2020-12-05 13:15 ` Varghese, Vipin
2020-12-04 7:51 ` [dpdk-dev] [PATCH 3/5] app/test-pmd: " Hideyuki Yamashita
2020-12-05 13:04 ` Varghese, Vipin
2020-12-04 7:51 ` [dpdk-dev] [PATCH 4/5] docs: add description of apistats parameter into proc-info Hideyuki Yamashita
2020-12-05 13:02 ` Varghese, Vipin
2020-12-07 9:48 ` Pattan, Reshma
2020-12-04 7:51 ` [dpdk-dev] [PATCH 5/5] librte_ethdev: add to use apistats Hideyuki Yamashita
2020-12-05 13:01 ` Varghese, Vipin
2020-12-06 17:47 ` Stephen Hemminger
2020-12-24 6:33 ` Hideyuki Yamashita
2020-12-07 12:38 ` Ananyev, Konstantin
2020-12-22 2:48 ` Hideyuki Yamashita
2020-12-22 2:50 ` Hideyuki Yamashita
2020-12-22 9:04 ` Morten Brørup [this message]
2020-12-22 13:05 ` Ananyev, Konstantin
2020-12-04 9:09 ` [dpdk-dev] [PATCH 0/5] add apistats function Ferruh Yigit
2020-12-04 10:20 ` David Hunt
2020-12-05 13:23 ` Varghese, Vipin
2020-12-24 6:43 ` Hideyuki Yamashita
2020-12-24 12:35 ` Varghese, Vipin
2020-12-22 2:16 ` Hideyuki Yamashita
2020-12-07 9:46 ` Thomas Monjalon
2020-12-22 2:22 ` Hideyuki Yamashita
2021-02-22 15:10 ` Ferruh Yigit
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=98CBD80474FA8B44BF855DF32C47DC35C614ED@smartserver.smartshare.dk \
--to=mb@smartsharesystems.com \
--cc=anatoly.burakov@intel.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=konstantin.ananyev@intel.com \
--cc=mdr@ashroe.eu \
--cc=nhorman@tuxdriver.com \
--cc=thomas@monjalon.net \
--cc=yamashita.hideyuki@ntt-tx.co.jp \
/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).