netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jakub Kicinski <kuba@kernel.org>
To: Nick Child <nnac123@linux.ibm.com>
Cc: netdev@vger.kernel.org, bjking1@linux.ibm.com,
	haren@linux.ibm.com, ricklind@us.ibm.com
Subject: Re: [PATCH net-next v2 7/7] ibmvnic: Perform tx CSO during send scrq direct
Date: Wed, 7 Aug 2024 08:08:38 -0700	[thread overview]
Message-ID: <20240807080838.289900bd@kernel.org> (raw)
In-Reply-To: <20240806193706.998148-8-nnac123@linux.ibm.com>

On Tue,  6 Aug 2024 14:37:06 -0500 Nick Child wrote:
> 1. NIC does checksum w headers, safely use send_subcrq_indirect:
>   - Packet rate: 631k txs
>   - Trace data:
>     ibmvnic_xmit = 44344685.87 us / 6234576 hits = AVG 7.11 us
>     ibmvnic_tx_scrq_flush = 33040649.69 us / 5638441 hits = AVG 5.86 us
>     send_subcrq_indirect = 37438922.24 us / 6030859 hits = AVG 6.21 us
>     skb_checksum_help = 4.07 us / 2 hits = AVG 2.04 us
>      ^ Notice hits, tracing this just for reassurance
> 
> 2. NIC does checksum w/o headers, dangerously use send_subcrq_direct:
>   - Packet rate: 831k txs
>   - Trace data:
>     ibmvnic_xmit = 48940092.29 us / 8187630 hits = AVG 5.98 us
>     ibmvnic_tx_scrq_flush = 31141879.57 us / 7948960 hits = AVG 3.92 us
>     send_subcrq_indirect = 8412506.03 us / 728781 hits = AVG 11.54
>      ^ notice hits is much lower
>     skb_checksum_help = 2.03 us / 1 hits = AVG 2.03
> 
> 3. driver does checksum, safely use send_subcrq_direct (THIS PATCH):
>   - Packet rate: 829k txs
>   - Trace data:
>     ibmvnic_xmit = 56696077.63 us / 8066168 hits = AVG 7.03 us
>     ibmvnic_tx_scrq_flush = 30219545.55 us / 7782409 hits = AVG 3.88 us
>     send_subcrq_indirect = 8638326.44 us / 763693 hits = AVG 11.31 us
>     skb_checksum_help = 8587456.16 us / 7526072 hits = AVG 1.14 us

Thanks for the numbers!

I presume the numbers are inclusive of all callees? It may be worth
while making ibmvnic_xmit more prominent, maybe indent the rest?
My first instinct was to add the AVGs in my head, and because of
different hit counts that's rather misguided.

      reply	other threads:[~2024-08-07 15:08 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-08-06 19:36 [PATCH net-next v2 0/7] ibmvnic rr patchset Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 1/7] ibmvnic: Only replenish rx pool when resources are getting low Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 2/7] ibmvnic: Use header len helper functions on tx Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 3/7] ibmvnic: Reduce memcpys in tx descriptor generation Nick Child
2024-08-07 13:16   ` Simon Horman
2024-08-06 19:37 ` [PATCH net-next v2 4/7] ibmvnic: Remove duplicate memory barriers in tx Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 5/7] ibmvnic: Introduce send sub-crq direct Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 6/7] ibmvnic: Only record tx completed bytes once per handler Nick Child
2024-08-06 19:37 ` [PATCH net-next v2 7/7] ibmvnic: Perform tx CSO during send scrq direct Nick Child
2024-08-07 15:08   ` Jakub Kicinski [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=20240807080838.289900bd@kernel.org \
    --to=kuba@kernel.org \
    --cc=bjking1@linux.ibm.com \
    --cc=haren@linux.ibm.com \
    --cc=netdev@vger.kernel.org \
    --cc=nnac123@linux.ibm.com \
    --cc=ricklind@us.ibm.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).