From: Ferruh Yigit <ferruh.yigit@intel.com>
To: Viacheslav Ovsiienko <viacheslavo@mellanox.com>, dev@dpdk.org
Cc: thomas@monjalon.net, bernard.iremonger@intel.com
Subject: Re: [dpdk-dev] [PATCH v2 3/3] app/testpmd: qualify profiling statistics on burst size
Date: Thu, 9 Apr 2020 13:03:41 +0100 [thread overview]
Message-ID: <56019b21-8578-dae1-5f65-485bac0a9643@intel.com> (raw)
In-Reply-To: <1584625851-10291-4-git-send-email-viacheslavo@mellanox.com>
On 3/19/2020 1:50 PM, Viacheslav Ovsiienko wrote:
> The execution time of rx/tx burst routine depends on the burst size.
> It would be meaningful to research this dependency, the patch
> provides an extra profiling data per rx/tx burst size.
>
> Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
<...>
> @@ -1568,6 +1568,23 @@ struct extmem_param {
> }
> printf(" + %d%% of %d pkts + %d%% of others]\n",
> burst_percent[1], (int) pktnb_stats[1], burst_percent[2]);
> +#ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
> + if (!(fwdprof_flags & (nrx_tx ? RECORD_CORE_CYCLES_TX
> + : RECORD_CORE_CYCLES_RX)))
> + return;
> + for (nb_pkt = 0; nb_pkt < MAX_PKT_BURST; nb_pkt++) {
> + nb_burst = nrx_tx ? pbs->pkt_retry_spread[nb_pkt]
> + : pbs->pkt_burst_spread[nb_pkt];
> + if (nb_burst == 0)
> + continue;
Why "nb_burst == 0" excluded, wouldn't be interesting to see number of calls
with 0 Rx/Tx and cycles spent there?
> + printf(" CPU cycles/%u packet burst=%u (total cycles="
> + "%"PRIu64" / total %s bursts=%u)\n",
> + (unsigned int)nb_pkt,
> + (unsigned int)(pbs->pkt_ticks_spread[nb_pkt] / nb_burst),
> + pbs->pkt_ticks_spread[nb_pkt],
> + nrx_tx ? "TX" : "RX", nb_burst);
> + }
> +#endif
> }
> #endif /* RTE_TEST_PMD_RECORD_BURST_STATS */
>
> @@ -1601,8 +1618,8 @@ struct extmem_param {
> }
>
> #ifdef RTE_TEST_PMD_RECORD_BURST_STATS
> - pkt_burst_stats_display("RX", &fs->rx_burst_stats);
> - pkt_burst_stats_display("TX", &fs->tx_burst_stats);
> + pkt_burst_stats_display(false, &fs->rx_burst_stats);
> + pkt_burst_stats_display(true, &fs->tx_burst_stats);
Instead of this true/false, I would be OK to some duplication and have explicit
Rx/Tx, I believe it would be more clear that way, but no strong opinion on it.
<...>
> +++ b/app/test-pmd/testpmd.h
> @@ -89,6 +89,10 @@ enum {
> */
> struct pkt_burst_stats {
> unsigned int pkt_burst_spread[MAX_PKT_BURST];
> +#ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
> + unsigned int pkt_retry_spread[MAX_PKT_BURST];
Isn't this keep the value of all Tx burst count, why named as 'retry'?
next prev parent reply other threads:[~2020-04-09 12:03 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-06-26 12:48 [dpdk-dev] [PATCH] app/testpmd: add profiling for Rx/Tx burst routines Viacheslav Ovsiienko
2019-06-26 12:57 ` Bruce Richardson
2019-06-26 13:19 ` Slava Ovsiienko
2019-06-26 13:21 ` Bruce Richardson
2019-06-27 4:48 ` Slava Ovsiienko
2019-06-28 13:45 ` Iremonger, Bernard
2019-06-28 14:20 ` Bruce Richardson
2019-07-01 4:57 ` Slava Ovsiienko
2019-07-01 8:15 ` Bruce Richardson
2019-09-30 12:32 ` Yigit, Ferruh
2020-03-19 13:50 ` [dpdk-dev] [PATCH v2 0/3] app/testpmd: qualify Rx/Tx profiling data on burst size Viacheslav Ovsiienko
2020-03-19 13:50 ` [dpdk-dev] [PATCH v2 1/3] app/testpmd: add profiling flags set command Viacheslav Ovsiienko
2020-04-02 11:15 ` Thomas Monjalon
2020-04-09 11:56 ` Ferruh Yigit
2020-04-13 7:56 ` Slava Ovsiienko
2020-04-13 12:23 ` Thomas Monjalon
2020-04-14 9:07 ` Ferruh Yigit
2020-03-19 13:50 ` [dpdk-dev] [PATCH v2 2/3] app/testpmd: gather Rx and Tx routines profiling Viacheslav Ovsiienko
2020-04-02 11:20 ` Thomas Monjalon
2020-04-02 11:23 ` Slava Ovsiienko
2020-03-19 13:50 ` [dpdk-dev] [PATCH v2 3/3] app/testpmd: qualify profiling statistics on burst size Viacheslav Ovsiienko
2020-03-20 6:13 ` Jerin Jacob
2020-04-09 11:46 ` Ferruh Yigit
2020-04-09 12:49 ` Jerin Jacob
2020-03-20 16:03 ` Andrzej Ostruszka
2020-04-02 11:21 ` Thomas Monjalon
2020-04-09 12:03 ` Ferruh Yigit [this message]
2020-04-09 12:09 ` Thomas Monjalon
2020-04-02 11:13 ` [dpdk-dev] [PATCH v2 0/3] app/testpmd: qualify Rx/Tx profiling data " 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=56019b21-8578-dae1-5f65-485bac0a9643@intel.com \
--to=ferruh.yigit@intel.com \
--cc=bernard.iremonger@intel.com \
--cc=dev@dpdk.org \
--cc=thomas@monjalon.net \
--cc=viacheslavo@mellanox.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).