From: Jerin Jacob <jerinjacobk@gmail.com>
To: Pavan Nikhilesh <pbhagavatula@marvell.com>
Cc: Jerin Jacob <jerinj@marvell.com>,
Shijith Thotton <sthotton@marvell.com>, dpdk-dev <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 2/2] event/cnxk: add Rx event vector fastpath
Date: Sun, 13 Jun 2021 18:44:49 +0530 [thread overview]
Message-ID: <CALBAE1Mu6m=dhbPKy7ZHKyfeiDqw--dLDTAh0TY7VXOYT6CeTw@mail.gmail.com> (raw)
In-Reply-To: <20210524130617.1621-2-pbhagavatula@marvell.com>
On Mon, May 24, 2021 at 6:36 PM <pbhagavatula@marvell.com> wrote:
>
> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
> Add Rx event vector fastpath to convert HW defined metadata into
> rte_mbuf and rte_event_vector.
>
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
> ---
> drivers/event/cnxk/cn10k_worker.h | 50 +++++++++++++++++++++++++++++++
> 1 file changed, 50 insertions(+)
>
> diff --git a/drivers/event/cnxk/cn10k_worker.h b/drivers/event/cnxk/cn10k_worker.h
> index b3f71202ad..8c2cd72873 100644
> --- a/drivers/event/cnxk/cn10k_worker.h
> +++ b/drivers/event/cnxk/cn10k_worker.h
> @@ -5,6 +5,8 @@
> #ifndef __CN10K_WORKER_H__
> #define __CN10K_WORKER_H__
>
> +#include <rte_vect.h>
> +
> #include "cnxk_ethdev.h"
> #include "cnxk_eventdev.h"
> #include "cnxk_worker.h"
> @@ -101,6 +103,44 @@ cn10k_wqe_to_mbuf(uint64_t wqe, const uint64_t mbuf, uint8_t port_id,
> mbuf_init.value, flags);
> }
>
> +static __rte_always_inline void
> +cn10k_process_vwqe(uintptr_t vwqe, uint16_t port_id, const uint32_t flags,
> + void *lookup_mem)
> +{
> + union mbuf_initializer mbuf_init = {
> + .fields = {.data_off = RTE_PKTMBUF_HEADROOM,
> + .refcnt = 1,
> + .nb_segs = 1,
> + .port = port_id},
> + };
If there is performance improvement move mbuf_init population to slow
path and store it
in HWS software structure or so.
> + struct rte_event_vector *vec;
> + uint16_t nb_mbufs, non_vec;
> + uint64_t **wqe;
> +
> + vec = (struct rte_event_vector *)vwqe;
> + wqe = vec->u64s;
> +
> + nb_mbufs = RTE_ALIGN_FLOOR(vec->nb_elem, NIX_DESCS_PER_LOOP);
> + nb_mbufs = cn10k_nix_recv_pkts_vector(&mbuf_init.value, vec->mbufs,
> + nb_mbufs, flags | NIX_RX_VWQE_F,
> + lookup_mem);
> + wqe += nb_mbufs;
> + non_vec = vec->nb_elem - nb_mbufs;
> +
> + while (non_vec) {
> + struct nix_cqe_hdr_s *cqe = (struct nix_cqe_hdr_s *)wqe[0];
> + struct rte_mbuf *mbuf;
> +
> + mbuf = (struct rte_mbuf *)((char *)cqe -
> + sizeof(struct rte_mbuf));
> + cn10k_nix_cqe_to_mbuf(cqe, cqe->tag, mbuf, lookup_mem,
> + mbuf_init.value, flags);
> + wqe[0] = (uint64_t *)mbuf;
> + non_vec--;
> + wqe++;
> + }
> +}
> +
> static __rte_always_inline uint16_t
> cn10k_sso_hws_get_work(struct cn10k_sso_hws *ws, struct rte_event *ev,
> const uint32_t flags, void *lookup_mem)
> @@ -141,6 +181,16 @@ cn10k_sso_hws_get_work(struct cn10k_sso_hws *ws, struct rte_event *ev,
> gw.u64[0] & 0xFFFFF, flags,
> lookup_mem);
> gw.u64[1] = mbuf;
> + } else if (CNXK_EVENT_TYPE_FROM_TAG(gw.u64[0]) ==
> + RTE_EVENT_TYPE_ETHDEV_VECTOR) {
> + uint8_t port = CNXK_SUB_EVENT_FROM_TAG(gw.u64[0]);
> + __uint128_t vwqe_hdr = *(__uint128_t *)gw.u64[1];
> +
> + vwqe_hdr = ((vwqe_hdr >> 64) & 0xFFF) | BIT_ULL(31) |
> + ((vwqe_hdr & 0xFFFF) << 48) |
> + ((uint64_t)port << 32);
> + *(uint64_t *)gw.u64[1] = (uint64_t)vwqe_hdr;
> + cn10k_process_vwqe(gw.u64[1], port, flags, lookup_mem);
> }
> }
>
> --
> 2.17.1
>
next prev parent reply other threads:[~2021-06-13 13:15 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-05-24 13:06 [dpdk-dev] [PATCH 1/2] event/cnxk: add Rx adapter vector support pbhagavatula
2021-05-24 13:06 ` [dpdk-dev] [PATCH 2/2] event/cnxk: add Rx event vector fastpath pbhagavatula
2021-06-13 13:14 ` Jerin Jacob [this message]
2021-06-13 13:18 ` [dpdk-dev] [PATCH 1/2] event/cnxk: add Rx adapter vector support Jerin Jacob
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='CALBAE1Mu6m=dhbPKy7ZHKyfeiDqw--dLDTAh0TY7VXOYT6CeTw@mail.gmail.com' \
--to=jerinjacobk@gmail.com \
--cc=dev@dpdk.org \
--cc=jerinj@marvell.com \
--cc=pbhagavatula@marvell.com \
--cc=sthotton@marvell.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).