From: "Morten Brørup" <mb@smartsharesystems.com>
To: <mattias.ronnblom@ericsson.com>
Cc: <olivier.matz@6wind.com>, <andrew.rybchenko@oktetlabs.ru>,
<dev@dpdk.org>, <thomas@monjalon.net>,
<david.marchand@redhat.com>, <honnappa.nagarahalli@arm.com>,
<konstantin.v.ananyev@yandex.ru>, <bruce.richardson@intel.com>
Subject: RE: [PATCH] clarify purpose of empty cache lines
Date: Tue, 5 Sep 2023 08:41:08 +0200 [thread overview]
Message-ID: <98CBD80474FA8B44BF855DF32C47DC35D87B70@smartserver.smartshare.dk> (raw)
In-Reply-To: <20230904084349.12044-1-mb@smartsharesystems.com>
> From: Morten Brørup [mailto:mb@smartsharesystems.com]
> Sent: Monday, 4 September 2023 10.44
>
> This patch introduces the generic RTE_CACHE_GUARD macro into the EAL, and
> replaces vaguely described empty cache lines in the rte_ring structure
> with this macro.
>
> Although the implementation of the rte_ring structure assumes that the
> hardware speculatively prefetches 1 cache line, this number can be changed
> at build time by modifying RTE_CACHE_GUARD_LINES in rte_config.h.
>
> The background and the RFC was discussed in this thread:
> http://inbox.dpdk.org/dev/98CBD80474FA8B44BF855DF32C47DC35D87B39@smartserver.s
> martshare.dk/
>
> Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> ---
> config/rte_config.h | 1 +
> lib/eal/include/rte_common.h | 13 +++++++++++++
> lib/ring/rte_ring_core.h | 6 +++---
> 3 files changed, 17 insertions(+), 3 deletions(-)
>
> diff --git a/config/rte_config.h b/config/rte_config.h
> index 400e44e3cf..cfdf787724 100644
> --- a/config/rte_config.h
> +++ b/config/rte_config.h
> @@ -37,6 +37,7 @@
> #define RTE_MAX_TAILQ 32
> #define RTE_LOG_DP_LEVEL RTE_LOG_INFO
> #define RTE_MAX_VFIO_CONTAINERS 64
> +#define RTE_CACHE_GUARD_LINES 1
>
> /* bsd module defines */
> #define RTE_CONTIGMEM_MAX_NUM_BUFS 64
> diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
> index 771c70f2c8..daf1866a32 100644
> --- a/lib/eal/include/rte_common.h
> +++ b/lib/eal/include/rte_common.h
> @@ -527,6 +527,19 @@ rte_is_aligned(const void * const __rte_restrict ptr,
> const unsigned int align)
> /** Force minimum cache line alignment. */
> #define __rte_cache_min_aligned __rte_aligned(RTE_CACHE_LINE_MIN_SIZE)
>
> +#define _RTE_CACHE_GUARD_HELPER2(unique) \
> + char cache_guard_ ## unique[RTE_CACHE_LINE_SIZE *
> RTE_CACHE_GUARD_LINES] \
> + __rte_cache_aligned
> +#define _RTE_CACHE_GUARD_HELPER1(unique) _RTE_CACHE_GUARD_HELPER2(unique)
> +/**
> + * Empty cache lines, to guard against false sharing-like effects
> + * on systems with a next-N-lines hardware prefetcher.
> + *
> + * Use as spacing between data accessed by different lcores,
> + * to prevent cache thrashing on hardware with speculative prefetching.
> + */
> +#define RTE_CACHE_GUARD _RTE_CACHE_GUARD_HELPER1(__COUNTER__)
> +
> /*********** PA/IOVA type definitions ********/
>
> /** Physical address */
> diff --git a/lib/ring/rte_ring_core.h b/lib/ring/rte_ring_core.h
> index d1e59bf9ad..327fdcf28f 100644
> --- a/lib/ring/rte_ring_core.h
> +++ b/lib/ring/rte_ring_core.h
> @@ -126,7 +126,7 @@ struct rte_ring {
> uint32_t mask; /**< Mask (size-1) of ring. */
> uint32_t capacity; /**< Usable size of ring */
>
> - char pad0 __rte_cache_aligned; /**< empty cache line */
> + RTE_CACHE_GUARD;
>
> /** Ring producer status. */
> union {
> @@ -135,7 +135,7 @@ struct rte_ring {
> struct rte_ring_rts_headtail rts_prod;
> } __rte_cache_aligned;
>
> - char pad1 __rte_cache_aligned; /**< empty cache line */
> + RTE_CACHE_GUARD;
>
> /** Ring consumer status. */
> union {
> @@ -144,7 +144,7 @@ struct rte_ring {
> struct rte_ring_rts_headtail rts_cons;
> } __rte_cache_aligned;
>
> - char pad2 __rte_cache_aligned; /**< empty cache line */
> + RTE_CACHE_GUARD;
> };
>
> #define RING_F_SP_ENQ 0x0001 /**< The default enqueue is "single-producer".
> */
> --
> 2.17.1
I asked Mattias Rönnblom in the RFC discussion thread to ACK this patch, which he then did [1], so I'm copying his ACK here for patchwork to see.
[1]: http://inbox.dpdk.org/dev/66b862bc-ca22-02da-069c-a82c2d52ed49@lysator.liu.se/
Acked-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>
next prev parent reply other threads:[~2023-09-05 6:41 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-09-04 8:43 Morten Brørup
2023-09-04 9:12 ` Bruce Richardson
2023-10-11 15:59 ` Thomas Monjalon
2023-09-05 6:41 ` Morten Brørup [this message]
2023-10-11 23:43 ` Thomas Monjalon
2023-10-12 6:00 ` Morten Brørup
2023-10-12 6:07 ` Morten Brørup
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=98CBD80474FA8B44BF855DF32C47DC35D87B70@smartserver.smartshare.dk \
--to=mb@smartsharesystems.com \
--cc=andrew.rybchenko@oktetlabs.ru \
--cc=bruce.richardson@intel.com \
--cc=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=honnappa.nagarahalli@arm.com \
--cc=konstantin.v.ananyev@yandex.ru \
--cc=mattias.ronnblom@ericsson.com \
--cc=olivier.matz@6wind.com \
--cc=thomas@monjalon.net \
/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).