From mboxrd@z Thu Jan 1 00:00:00 1970
Return-Path: <dev-bounces@dpdk.org>
Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124])
by inbox.dpdk.org (Postfix) with ESMTP id B6C5A43B0D;
Tue, 13 Feb 2024 19:48:20 +0100 (CET)
Received: from mails.dpdk.org (localhost [127.0.0.1])
by mails.dpdk.org (Postfix) with ESMTP id 533FC42E1B;
Tue, 13 Feb 2024 19:48:20 +0100 (CET)
Received: from linux.microsoft.com (linux.microsoft.com [13.77.154.182])
by mails.dpdk.org (Postfix) with ESMTP id E38B442E10
for <dev@dpdk.org>; Tue, 13 Feb 2024 19:48:18 +0100 (CET)
Received: by linux.microsoft.com (Postfix, from userid 1086)
id 2C84A20B2000; Tue, 13 Feb 2024 10:48:18 -0800 (PST)
DKIM-Filter: OpenDKIM Filter v2.11.0 linux.microsoft.com 2C84A20B2000
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linux.microsoft.com;
s=default; t=1707850098;
bh=uiMjfiY4nAxd5azDWzmHOG9RRwxAchNYv47sa0w2qsQ=;
h=Date:From:To:Cc:Subject:References:In-Reply-To:From;
b=AtZ56ge0++XHzzbCx9qqfU+WoVvJhMrTjBss1IrEnfFBtCbgB3O63G9h8KTHZCyvb
21Vg4U6KZDYmCHmOQ3zAu4KZ9JF8InjPb99g839yKmArBPRqYFcjGEEDZW2iumeCAI
sUIz/k53B43x+Asltv/VuhSyn9fBLEhvuYvKqUXI=
Date: Tue, 13 Feb 2024 10:48:18 -0800
From: Tyler Retzlaff <roretzla@linux.microsoft.com>
To: Morten =?iso-8859-1?Q?Br=F8rup?= <mb@smartsharesystems.com>
Cc: dev@dpdk.org, Andrew Boyer <andrew.boyer@amd.com>,
Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>,
Bruce Richardson <bruce.richardson@intel.com>,
Chenbo Xia <chenbox@nvidia.com>,
Konstantin Ananyev <konstantin.v.ananyev@yandex.ru>,
Maxime Coquelin <maxime.coquelin@redhat.com>
Subject: Re: [PATCH v2] mbuf: replace GCC marker extension with C11 anonymous
unions
Message-ID: <20240213184818.GA20546@linuxonhyperv3.guj3yctzbm1etfxqx2vob5hsef.xx.internal.cloudapp.net>
References: <1706657173-26166-2-git-send-email-roretzla@linux.microsoft.com>
<1707806741-29694-1-git-send-email-roretzla@linux.microsoft.com>
<1707806741-29694-2-git-send-email-roretzla@linux.microsoft.com>
<98CBD80474FA8B44BF855DF32C47DC35E9F20F@smartserver.smartshare.dk>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <98CBD80474FA8B44BF855DF32C47DC35E9F20F@smartserver.smartshare.dk>
User-Agent: Mutt/1.5.21 (2010-09-15)
X-BeenThere: dev@dpdk.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: DPDK patches and discussions <dev.dpdk.org>
List-Unsubscribe: <https://mails.dpdk.org/options/dev>,
<mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://mails.dpdk.org/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <https://mails.dpdk.org/listinfo/dev>,
<mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
On Tue, Feb 13, 2024 at 05:58:21PM +0100, Morten Brørup wrote:
> > From: Tyler Retzlaff [mailto:roretzla@linux.microsoft.com]
> > Sent: Tuesday, 13 February 2024 07.46
> >
> > Replace the use of RTE_MARKER<x> with C11 anonymous unions to improve
> > code portability between toolchains.
>
> How about combining the cacheline 0 marker and padding, like this:
this seems like a good suggestion i will evaluate it. at least it gets
rid of all the extra nesting if there are no unforseen problems.
>
> struct rte_mbuf {
> - RTE_MARKER cacheline0;
> + union {
> + char cacheline0[RTE_CACHE_LINE_MIN_SIZE];
>
> + struct {
> - void *buf_addr; /**< Virtual address of segment buffer. */
> + void *buf_addr; /**< Virtual address of segment buffer. */
> #if RTE_IOVA_IN_MBUF
>
>
> You could do the same with the cacheline1 marker:
yeah, i wondered if i should. i'll do it since it does seem more
consistent to just pad out both cachelines explicitly instead of just
doing all but the last.
we probably don't need to align struct rte_mbuf type if we do since it
will cause it to be naturally aligned to RTE_CACHE_LINE_MIN_SIZE.
>
> /* second cache line - fields only used in slow path or on TX */
> - RTE_MARKER cacheline1 __rte_cache_min_aligned;
> + union {
> + char cacheline1[RTE_CACHE_LINE_MIN_SIZE];
>
> + struct {
> #if RTE_IOVA_IN_MBUF
> - /**
> - * Next segment of scattered packet. Must be NULL in the last
> - * segment or in case of non-segmented packet.
> - */
> - struct rte_mbuf *next;
> + /**
> + * Next segment of scattered packet. Must be NULL in the last
> + * segment or in case of non-segmented packet.
> + */
> + struct rte_mbuf *next;
> #else
>
>
> It also avoids the weird union between cacheline0 and buf_addr at the beginning of the structure, and between cacheline1 and next/dynfield2 at the beginning of the second cache line.
>
> And then you can omit the pad_cacheline0 array at the end of the first part of the structure.
>
>
> BTW: char is a weaker type than uint8_t - i.e. it is easier to cast to another type.
> It might be a personal preference, but I would use char instead of uint8_t for the padding array.
noted, i'll make the change.
thanks!