From: "Kundapura, Ganapati" <ganapati.kundapura@intel.com>
To: Jerin Jacob <jerinjacobk@gmail.com>, dpdk-dev <dev@dpdk.org>,
"Jayatheerthan, Jay" <jay.jayatheerthan@intel.com>
Cc: "Naga Harish K, S V" <s.v.naga.harish.k@intel.com>,
"Kundapura, Ganapati" <ganapati.kundapura@intel.com>
Subject: [dpdk-dev] RFC V2: Enhancements to Rx adapter for DPDK 21.11
Date: Fri, 3 Sep 2021 12:20:56 +0000 [thread overview]
Message-ID: <CO1PR11MB48823A11EA86308F6B299C1687CF9@CO1PR11MB4882.namprd11.prod.outlook.com> (raw)
Hi dpdk-dev,
This is the updated RFC for enhancements to Rx adapter that was
communicated earlier.
Please find the details below.
(1) Configure Rx event buffer size during adapter instance creation
Add new api to specify the Rx event buffer size at adapter creation and
allocate the event buffer with size specified.
Earlier the RFC mentioned configuration of Rx event buffer size at run time.
Changing the buffer size at run time can affect the traffic and hence
this is being updated to adapter creation time.
(2) Allocate event buffer per RxQ.
(1) is event buffer per adapter which is same for all Rx queueus.
In this, event buffer is allocated per Rx queue.
rte_event_eth_rx_adapter_queue_add() is proposed to be updated to take the size specified in
rte_event_eth_rx_adapter_queue_conf structure to allocate event buffer per RxQ.
(3) Change packet enqueue buffer in Rx adapter to circular buffer
Rx adapter uses memmove() to move unprocessed events to the begining
of packet enqueue buffer which consumes good amount of CPU cycles.
This patch is already posted for review. This is going to superseded by
(1), (2) and (3) together.
(4) Add API to retrieve the Rx queue info
Rx queue info containining flags for handling received packets,
event queue identifier, schedular type, event priority,
polling frequence of the receive queue and flow identifier
(5) Add adapter_stats cli to retrive Rx/Tx adapter stats and rxq info
This cli displays Rx and Tx adapter stats containing recieved packet count,
eventdev enqueue count, enqueue retry count, event buffer size, queue poll count,
transmitted packet count, packet dropped count, transmit fail count etc and rx queue info.
We are investigating on how to add this as part of telemetry.
(6) Update Rx timestamp in mbuf using mbuf dynamic field
Add support to register timestamp dynamic field in mbuf
Update the timestamp in mbuf for each packet before eventdev enqueue
We look forward for your feedback on this proposal.
Thanks,
Ganapati
reply other threads:[~2021-09-03 12:21 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
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=CO1PR11MB48823A11EA86308F6B299C1687CF9@CO1PR11MB4882.namprd11.prod.outlook.com \
--to=ganapati.kundapura@intel.com \
--cc=dev@dpdk.org \
--cc=jay.jayatheerthan@intel.com \
--cc=jerinjacobk@gmail.com \
--cc=s.v.naga.harish.k@intel.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).