From: David Marchand <david.marchand@redhat.com>
To: Stephen Hemminger <stephen@networkplumber.org>
Cc: dev <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v7] eal: make lcore_config private
Date: Wed, 2 Oct 2019 10:15:34 +0200 [thread overview]
Message-ID: <CAJFAV8xdaE5bFTOXz2qRhkL5kd_HR3vTACkbpNLa0QfgQ2fEHA@mail.gmail.com> (raw)
In-Reply-To: <20190925161013.3656-1-stephen@networkplumber.org>
Thanks for working on this.
On Wed, Sep 25, 2019 at 6:10 PM Stephen Hemminger
<stephen@networkplumber.org> wrote:
>
> The internal structure of lcore_config is no longer be part of
> visible API/ABI. Make it private to EAL.
>
> Rearrange and resize the fields in the structure so it takes
> less memory (and cache footprint).
This patch is missing the release notes update.
>
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
> v7 - add eal_private.h to windows
>
> lib/librte_eal/common/eal_common_launch.c | 2 ++
> lib/librte_eal/common/eal_private.h | 22 +++++++++++++++++++++
> lib/librte_eal/common/include/rte_lcore.h | 24 -----------------------
> lib/librte_eal/common/rte_service.c | 2 ++
> lib/librte_eal/rte_eal_version.map | 1 -
> lib/librte_eal/windows/eal/eal_thread.c | 1 +
> 6 files changed, 27 insertions(+), 25 deletions(-)
>
> diff --git a/lib/librte_eal/common/eal_common_launch.c b/lib/librte_eal/common/eal_common_launch.c
> index fe0ba3f0d617..cf52d717f68e 100644
> --- a/lib/librte_eal/common/eal_common_launch.c
> +++ b/lib/librte_eal/common/eal_common_launch.c
> @@ -15,6 +15,8 @@
> #include <rte_per_lcore.h>
> #include <rte_lcore.h>
>
> +#include "eal_private.h"
> +
> /*
> * Wait until a lcore finished its job.
> */
> diff --git a/lib/librte_eal/common/eal_private.h b/lib/librte_eal/common/eal_private.h
> index 798ede553b21..25e80547904f 100644
> --- a/lib/librte_eal/common/eal_private.h
> +++ b/lib/librte_eal/common/eal_private.h
> @@ -10,6 +10,28 @@
> #include <stdio.h>
>
> #include <rte_dev.h>
> +#include <rte_lcore.h>
> +
> +/**
> + * Structure storing internal configuration (per-lcore)
> + */
> +struct lcore_config {
> + uint32_t core_id; /**< core number on socket for this lcore */
> + uint32_t core_index; /**< relative index, starting from 0 */
> + uint16_t socket_id; /**< physical socket id for this lcore */
> + uint8_t core_role; /**< role of core eg: OFF, RTE, SERVICE */
> + uint8_t detected; /**< true if lcore was detected */
> + volatile enum rte_lcore_state_t state; /**< lcore state */
> + rte_cpuset_t cpuset; /**< cpu set which the lcore affinity to */
> + pthread_t thread_id; /**< pthread identifier */
> + int pipe_master2slave[2]; /**< communication pipe with master */
> + int pipe_slave2master[2]; /**< communication pipe with master */
> + lcore_function_t * volatile f; /**< function to call */
> + void * volatile arg; /**< argument of function */
> + volatile int ret; /**< return value of function */
> +};
> +
> +extern struct lcore_config lcore_config[RTE_MAX_LCORE];
Everything but cpuset can fit in a cache line.
You could just move the cpuset field at the end of the structure and
change detected to uint8_t.
This gives the following layout:
struct lcore_config {
pthread_t thread_id; /* 0 8 */
int pipe_master2slave[2]; /* 8 8 */
int pipe_slave2master[2]; /* 16 8 */
volatile lcore_function_t * f; /* 24 8 */
volatile void * arg; /* 32 8 */
volatile int ret; /* 40 4 */
volatile enum rte_lcore_state_t state; /* 44 4 */
unsigned int socket_id; /* 48 4 */
unsigned int core_id; /* 52 4 */
int core_index; /* 56 4 */
uint8_t detected; /* 60 1 */
uint8_t core_role; /* 61 1 */
/* XXX 2 bytes hole, try to pack */
/* --- cacheline 1 boundary (64 bytes) --- */
rte_cpuset_t cpuset; /* 64 128 */
/* --- cacheline 3 boundary (192 bytes) --- */
/* size: 192, cachelines: 3, members: 13 */
/* sum members: 190, holes: 1, sum holes: 2 */
};
The resulting structure is only two bytes bigger than your proposal
and does not touch existing integer types (avoiding the risk of some
integer conversion on socket_id for example).
--
David Marchand
next prev parent reply other threads:[~2019-10-02 8:15 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-09-25 16:10 Stephen Hemminger
2019-10-02 8:15 ` David Marchand [this message]
2019-10-02 19:40 ` [dpdk-dev] [PATCH v8] " Stephen Hemminger
2019-10-22 9:05 ` David Marchand
2019-10-22 16:30 ` Stephen Hemminger
2019-10-22 16:49 ` David Marchand
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=CAJFAV8xdaE5bFTOXz2qRhkL5kd_HR3vTACkbpNLa0QfgQ2fEHA@mail.gmail.com \
--to=david.marchand@redhat.com \
--cc=dev@dpdk.org \
--cc=stephen@networkplumber.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).