DPDK patches and discussions
 help / color / mirror / Atom feed
From: Bruce Richardson <bruce.richardson@intel.com>
To: "Burakov, Anatoly" <anatoly.burakov@intel.com>
Cc: <dev@dpdk.org>
Subject: Re: [PATCH v4 23/25] net/intel: support wider x86 vectors for Rx rearm
Date: Thu, 5 Jun 2025 10:31:59 +0100	[thread overview]
Message-ID: <aEFkD1gYW3vAVZC4@bricha3-mobl1.ger.corp.intel.com> (raw)
In-Reply-To: <e9a7a3e8-2154-4c4f-9846-5d2241627d36@intel.com>

On Thu, Jun 05, 2025 at 11:29:11AM +0200, Burakov, Anatoly wrote:
> On 6/4/2025 4:59 PM, Bruce Richardson wrote:
> > On Fri, May 30, 2025 at 02:57:19PM +0100, Anatoly Burakov wrote:
> > > Currently, for 32-byte descriptor format, only SSE instruction set is
> > > supported. Add implementation for AVX2 and AVX512 instruction sets. Since
> > > we are using Rx descriptor definitions from common code, we can just use
> > > the generic descriptor definition, as we only ever write the first 16 bytes
> > > of it, and the layout is always the same for that part.
> > > 
> > > Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
> > > ---
> > > 
> > 
> > Like the idea. Feedback inline below.
> > 
> > /Bruce
> > 
> 
<snip>

> 
> > > -		/**
> > > -		 * merge 0 & 1, by casting 0 to 256-bit and inserting 1
> > > -		 * into the high lanes. Similarly for 2 & 3, and so on.
> > > -		 */
> > > -		const __m256i addr0_256 = _mm256_castsi128_si256(vaddr0);
> > > -		const __m256i addr2_256 = _mm256_castsi128_si256(vaddr2);
> > > -		const __m256i addr4_256 = _mm256_castsi128_si256(vaddr4);
> > > -		const __m256i addr6_256 = _mm256_castsi128_si256(vaddr6);
> > > +			const __m128i vaddr0 = _mm_loadu_si128((const __m128i *)&mb0->buf_addr);
> > > +			const __m128i vaddr1 = _mm_loadu_si128((const __m128i *)&mb1->buf_addr);
> > > +			const __m128i vaddr2 = _mm_loadu_si128((const __m128i *)&mb2->buf_addr);
> > > +			const __m128i vaddr3 = _mm_loadu_si128((const __m128i *)&mb3->buf_addr);
> > > -		const __m256i addr0_1 = _mm256_inserti128_si256(addr0_256, vaddr1, 1);
> > > -		const __m256i addr2_3 = _mm256_inserti128_si256(addr2_256, vaddr3, 1);
> > > -		const __m256i addr4_5 = _mm256_inserti128_si256(addr4_256, vaddr5, 1);
> > > -		const __m256i addr6_7 = _mm256_inserti128_si256(addr6_256, vaddr7, 1);
> > > +			reg0 = _ci_rxq_rearm_desc_avx512(vaddr0, zero, vaddr1, zero);
> > > +			reg1 = _ci_rxq_rearm_desc_avx512(vaddr2, zero, vaddr3, zero);
> > 
> > I can't help but thinking we can probably do a little better than this
> > merging in zeros using AVX-512 mask registers, e.g. using
> >   _mm256_maskz_broadcastq_epi64()  intrinsic, but it will be ok for now! :-)
> > 
> 
> You're welcome to submit patches, this is a very welcoming community!
> 
> (seriously though, I'll look into it)
> 
Great, but I also think this is fine for now if you want to keep it as-is.
We can do a post-rework optimization patchset after this goes in. Main
thing is just to ensure we don't see a perf regression after this work.

/Bruce

  reply	other threads:[~2025-06-05  9:32 UTC|newest]

Thread overview: 113+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-06 13:27 [PATCH v1 01/13] net/ixgbe: remove unused field in Rx queue struct Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 02/13] net/iavf: make IPsec stats dynamically allocated Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 03/13] net/ixgbe: create common Rx queue structure Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 04/13] net/i40e: use the " Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 05/13] net/ice: " Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 06/13] net/iavf: " Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 07/13] net/intel: generalize vectorized Rx rearm Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 08/13] net/i40e: use common Rx rearm code Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 09/13] net/iavf: " Anatoly Burakov
2025-05-06 13:27 ` [PATCH v1 10/13] net/ixgbe: " Anatoly Burakov
2025-05-06 13:28 ` [PATCH v1 11/13] net/intel: support wider x86 vectors for Rx rearm Anatoly Burakov
2025-05-06 13:28 ` [PATCH v1 12/13] net/intel: add common Rx mbuf recycle Anatoly Burakov
2025-05-06 13:28 ` [PATCH v1 13/13] net/intel: add common Tx " Anatoly Burakov
2025-05-12 10:58 ` [PATCH v2 01/13] net/ixgbe: remove unused field in Rx queue struct Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 02/13] net/iavf: make IPsec stats dynamically allocated Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 03/13] net/ixgbe: create common Rx queue structure Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 04/13] net/i40e: use the " Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 05/13] net/ice: " Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 06/13] net/iavf: " Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 07/13] net/intel: generalize vectorized Rx rearm Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 08/13] net/i40e: use common Rx rearm code Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 09/13] net/iavf: " Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 10/13] net/ixgbe: " Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 11/13] net/intel: support wider x86 vectors for Rx rearm Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 12/13] net/intel: add common Rx mbuf recycle Anatoly Burakov
2025-05-12 10:58   ` [PATCH v2 13/13] net/intel: add common Tx " Anatoly Burakov
2025-05-12 12:54 ` [PATCH v3 01/13] net/ixgbe: remove unused field in Rx queue struct Anatoly Burakov
2025-05-12 12:54   ` [PATCH v3 02/13] net/iavf: make IPsec stats dynamically allocated Anatoly Burakov
2025-05-14 16:39     ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 03/13] net/ixgbe: create common Rx queue structure Anatoly Burakov
2025-05-14 16:45     ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 04/13] net/i40e: use the " Anatoly Burakov
2025-05-14 16:52     ` Bruce Richardson
2025-05-15 11:09       ` Burakov, Anatoly
2025-05-15 12:55         ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 05/13] net/ice: " Anatoly Burakov
2025-05-14 16:56     ` Bruce Richardson
2025-05-23 11:16       ` Burakov, Anatoly
2025-05-12 12:54   ` [PATCH v3 06/13] net/iavf: " Anatoly Burakov
2025-05-15 10:59     ` Bruce Richardson
2025-05-15 11:11       ` Burakov, Anatoly
2025-05-15 12:57         ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 07/13] net/intel: generalize vectorized Rx rearm Anatoly Burakov
2025-05-15 10:56     ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 08/13] net/i40e: use common Rx rearm code Anatoly Burakov
2025-05-15 10:58     ` Bruce Richardson
2025-05-12 12:54   ` [PATCH v3 09/13] net/iavf: " Anatoly Burakov
2025-05-12 12:54   ` [PATCH v3 10/13] net/ixgbe: " Anatoly Burakov
2025-05-12 12:54   ` [PATCH v3 11/13] net/intel: support wider x86 vectors for Rx rearm Anatoly Burakov
2025-05-12 12:54   ` [PATCH v3 12/13] net/intel: add common Rx mbuf recycle Anatoly Burakov
2025-05-12 12:54   ` [PATCH v3 13/13] net/intel: add common Tx " Anatoly Burakov
2025-05-15 11:07     ` Bruce Richardson
2025-05-12 12:58   ` [PATCH v3 01/13] net/ixgbe: remove unused field in Rx queue struct Bruce Richardson
2025-05-14 16:32   ` Bruce Richardson
2025-05-15 11:15     ` Burakov, Anatoly
2025-05-15 12:58       ` Bruce Richardson
2025-05-30 13:56 ` [PATCH v4 00/25] Intel PMD drivers Rx cleanp Anatoly Burakov
2025-05-30 13:56   ` [PATCH v4 01/25] net/ixgbe: remove unused field in Rx queue struct Anatoly Burakov
2025-05-30 13:56   ` [PATCH v4 02/25] net/iavf: make IPsec stats dynamically allocated Anatoly Burakov
2025-05-30 13:56   ` [PATCH v4 03/25] net/ixgbe: match variable names to other drivers Anatoly Burakov
2025-06-03 15:54     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 04/25] net/i40e: match variable name " Anatoly Burakov
2025-06-03 15:56     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 05/25] net/ice: " Anatoly Burakov
2025-06-03 15:57     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 06/25] net/i40e: rename 16-byte descriptor define Anatoly Burakov
2025-06-03 15:58     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 07/25] net/ice: " Anatoly Burakov
2025-06-03 15:59     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 08/25] net/iavf: " Anatoly Burakov
2025-06-03 16:06     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 09/25] net/ixgbe: simplify vector PMD compilation Anatoly Burakov
2025-06-03 16:09     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 10/25] net/ixgbe: replace always-true check Anatoly Burakov
2025-06-03 16:15     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 11/25] net/ixgbe: clean up definitions Anatoly Burakov
2025-06-03 16:17     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 12/25] net/i40e: " Anatoly Burakov
2025-06-03 16:19     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 13/25] net/ice: " Anatoly Burakov
2025-06-03 16:20     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 14/25] net/iavf: " Anatoly Burakov
2025-06-03 16:21     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 15/25] net/ixgbe: create common Rx queue structure Anatoly Burakov
2025-06-03 16:45     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 16/25] net/i40e: use the " Anatoly Burakov
2025-06-03 16:57     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 17/25] net/ice: " Anatoly Burakov
2025-06-03 17:02     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 18/25] net/iavf: " Anatoly Burakov
2025-06-03 17:05     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 19/25] net/intel: generalize vectorized Rx rearm Anatoly Burakov
2025-06-04  9:32     ` Bruce Richardson
2025-06-04  9:43       ` Morten Brørup
2025-06-04  9:49         ` Bruce Richardson
2025-06-04 10:18           ` Morten Brørup
2025-05-30 13:57   ` [PATCH v4 20/25] net/i40e: use common Rx rearm code Anatoly Burakov
2025-06-04  9:33     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 21/25] net/iavf: " Anatoly Burakov
2025-06-04  9:34     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 22/25] net/ixgbe: " Anatoly Burakov
2025-06-04  9:40     ` Bruce Richardson
2025-06-05  9:22       ` Burakov, Anatoly
2025-05-30 13:57   ` [PATCH v4 23/25] net/intel: support wider x86 vectors for Rx rearm Anatoly Burakov
2025-06-04 12:32     ` Bruce Richardson
2025-06-04 14:59     ` Bruce Richardson
2025-06-05  9:29       ` Burakov, Anatoly
2025-06-05  9:31         ` Bruce Richardson [this message]
2025-06-05 10:09         ` Morten Brørup
2025-05-30 13:57   ` [PATCH v4 24/25] net/intel: add common Rx mbuf recycle Anatoly Burakov
2025-06-04 15:09     ` Bruce Richardson
2025-05-30 13:57   ` [PATCH v4 25/25] net/intel: add common Tx " Anatoly Burakov
2025-06-04 15:18     ` Bruce Richardson

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=aEFkD1gYW3vAVZC4@bricha3-mobl1.ger.corp.intel.com \
    --to=bruce.richardson@intel.com \
    --cc=anatoly.burakov@intel.com \
    --cc=dev@dpdk.org \
    /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).