DPDK patches and discussions
 help / color / mirror / Atom feed
* NVIDIA roadmap for 24.11
@ 2024-08-25 11:58 Maayan Kashani
  0 siblings, 0 replies; only message in thread
From: Maayan Kashani @ 2024-08-25 11:58 UTC (permalink / raw)
  To: dpdk-dev

[-- Attachment #1: Type: text/plain, Size: 2032 bytes --]

Please find below NVIDIA roadmap for 24.11 release:

A. rte_flow new APIs 
================= 

  1.  New action - Jump to a table index
Jump to a table index directly without doing the hash inside the matcher to improve the performance by reducing hops.
This feature enables distribution based on various fields, not subject to RSS only fields.

B. Net/mlx5 PMD updates 
===================== 

  1.  Optimize FDB domain tables​ 
Reduce the duplication of Tx / Rx tables in the FDB domain to optimize the performance and reduce memory
usage to allow room for more offloads (increase total scale). 

  1.  Hairpin statistics query​ 
Support per queue hairpin related counters, Rx miss, Tx error and queue drop, etc.​ 
This will provide more visibility for debug, telemetry, and monitoring purposes to the users​. 
This feature is supported from CX7/BF3 and above and on the VF. For CX6-Dx it supports port level stats and PF.

  1.  RTE FLOW backward compatible API​ 
Support RTE FLOW non-template API on top of template async API.  
Strating from ConnectX-8 NIC and BlueField-4 DPU,  RTE FLOW non-template API will not be supported.
This feature will allow the non-template API to be supported with template mode for interim
application translation and with functional template mode limitations.

  1.  Report detail reason of HWS rule insertion failure​
Provide detailed failure reason to the upper layer, e.g. table full, hash conflict, input error​​.
The application can handle based on the failures.​​
  
C. Test apps updates   
=================  

  1.  Support the changes in rte_flow listed above in testpmd.  

  1.  Support hairpin with 4 physical ports in testpmd. 
For Nic’s supporting more than dual ports, adjust the hairpin logic to allow use of all ports.  
Current testpmd only allows consecutive pair for hairpin.
This feature is intended to allow use of any pair. 


Regards,
Maayan Kashani


[-- Attachment #2: Type: text/html, Size: 10220 bytes --]

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2024-08-25 11:58 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-08-25 11:58 NVIDIA roadmap for 24.11 Maayan Kashani

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).