From: Serge Semin <fancer.lancer@gmail.com>
To: Alexander Fomichev <fomichev.ru@gmail.com>
Cc: ntb@lists.linux.dev, linux@yadro.com,
Jon Mason <jdmason@kudzu.us>, Dave Jiang <dave.jiang@intel.com>,
Allen Hubbe <allenbh@gmail.com>,
Guo Zhengkui <guozhengkui@vivo.com>,
Alexander Fomichev <a.fomichev@yadro.com>
Subject: Re: [PATCH v4 0/3] ntb_perf: add new 'latency' test set
Date: Sun, 21 Aug 2022 21:48:55 +0300 [thread overview]
Message-ID: <20220821184855.shajhgjq54ohctwc@mobilestation> (raw)
In-Reply-To: <20220812165544.26307-1-fomichev.ru@gmail.com>
On Fri, Aug 12, 2022 at 07:55:41PM +0300, Alexander Fomichev wrote:
> From: Alexander Fomichev <a.fomichev@yadro.com>
>
> The ntb_perf test provides a tool for NTB hardware performance
> evaluation. For software impact elimination the test uses a simple method
> (let's call it 'burst' mode), when the local system sends to the remote
> system a data set and counts time interval until hardware completion
> report, without the remote side confirming, nor data integrity check.
> The measured metric is a 'burst' throughput bandwidth of NTB connection.
>
> The patches extend ntb_perf with 3 new metrics:
> 1) Burst latency
> 2) Poll latency
> 3) Doorbell latency
>
> The resulting test set is fully backward compatible.
As before I think the latency test would be better to live in a
separate driver as I suggested here:
https://lore.kernel.org/ntb/20220622203648.jo6raa4h57g24el2@mobilestation/
The Messages/Spad+Doorbell-based communication part of the ntb_perf
driver could be moved into a separate module as a preparation patch.
The best approach would to create some kind of NTB communication
library, which could be used not only by the perf and latency tests,
but by the client drivers like NTB-transport, etc. Thus we'll finally
have a complete initialization solution for all types of the NTB
hardware: local and peer MW-based, Messages and Spad-based.
-Sergey
>
> Alexander Fomichev (3):
> ntb_perf: extend with burst latency measurement
> ntb_perf: extend with poll latency measurement
> ntb_perf: extend with doorbell latency measurement
>
> drivers/ntb/test/ntb_perf.c | 829 +++++++++++++++++++++++++++++++++++-
> 1 file changed, 813 insertions(+), 16 deletions(-)
>
> --
> 2.37.1
>
prev parent reply other threads:[~2022-08-21 18:48 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-12 16:55 [PATCH v4 0/3] ntb_perf: add new 'latency' test set Alexander Fomichev
2022-08-12 16:55 ` [PATCH v4 1/3] ntb_perf: extend with burst latency measurement Alexander Fomichev
2022-08-12 16:55 ` [PATCH v4 2/3] ntb_perf: extend with poll " Alexander Fomichev
2022-08-12 16:55 ` [PATCH v4 3/3] ntb_perf: extend with doorbell " Alexander Fomichev
2022-08-15 16:45 ` [PATCH v4 0/3] ntb_perf: add new 'latency' test set Dave Jiang
2022-08-16 17:56 ` Alexander Fomichev
2022-08-21 18:48 ` Serge Semin [this message]
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=20220821184855.shajhgjq54ohctwc@mobilestation \
--to=fancer.lancer@gmail.com \
--cc=a.fomichev@yadro.com \
--cc=allenbh@gmail.com \
--cc=dave.jiang@intel.com \
--cc=fomichev.ru@gmail.com \
--cc=guozhengkui@vivo.com \
--cc=jdmason@kudzu.us \
--cc=linux@yadro.com \
--cc=ntb@lists.linux.dev \
/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 read-only IMAP folder(s) and NNTP newsgroup(s).