From: Yongseok Koh <yskoh@mellanox.com>
To: "Nélio Laranjeiro" <nelio.laranjeiro@6wind.com>
Cc: Ferruh Yigit <ferruh.yigit@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>,
Adrien Mazarguil <adrien.mazarguil@6wind.com>
Subject: Re: [dpdk-dev] [PATCH v2 3/5] net/mlx5: use buffer address for LKEY search
Date: Mon, 3 Jul 2017 20:54:43 +0000 [thread overview]
Message-ID: <4771B0EC-2A1E-4A0C-86C5-231289F3E0F2@mellanox.com> (raw)
In-Reply-To: <20170703140605.GE18305@autoinstall.dev.6wind.com>
> On Jul 3, 2017, at 7:06 AM, Nélio Laranjeiro <nelio.laranjeiro@6wind.com> wrote:
>
> On Fri, Jun 30, 2017 at 12:23:31PM -0700, Yongseok Koh wrote:
>> When searching LKEY, if search key is mempool pointer, the 2nd cacheline
>> has to be accessed and it even requires to check whether a buffer is
>> indirect per every search. Instead, using address for search key can reduce
>> cycles taken. And caching the last hit entry is beneficial as well.
>>
>> Signed-off-by: Yongseok Koh <yskoh@mellanox.com>
>> ---
>> drivers/net/mlx5/mlx5_mr.c | 17 ++++++++++++++---
>> drivers/net/mlx5/mlx5_rxtx.c | 39 +++++++++++++++++++++------------------
>> drivers/net/mlx5/mlx5_rxtx.h | 4 +++-
>> drivers/net/mlx5/mlx5_txq.c | 3 +--
>> 4 files changed, 39 insertions(+), 24 deletions(-)
>>
>> diff --git a/drivers/net/mlx5/mlx5_mr.c b/drivers/net/mlx5/mlx5_mr.c
>> index 0a3638460..287335179 100644
>> --- a/drivers/net/mlx5/mlx5_mr.c
>> +++ b/drivers/net/mlx5/mlx5_mr.c
>> @@ -265,18 +266,28 @@ txq_mp2mr_iter(struct rte_mempool *mp, void *arg)
>> struct txq_mp2mr_mbuf_check_data data = {
>> .ret = 0,
>> };
>> + uintptr_t start;
>> + uintptr_t end;
>> unsigned int i;
>>
>> /* Register mempool only if the first element looks like a mbuf. */
>> if (rte_mempool_obj_iter(mp, txq_mp2mr_mbuf_check, &data) == 0 ||
>> data.ret == -1)
>> return;
>> + if (mlx5_check_mempool(mp, &start, &end) != 0) {
>> + ERROR("mempool %p: not virtually contiguous",
>> + (void *)mp);
>> + return;
>> + }
>> for (i = 0; (i != RTE_DIM(txq_ctrl->txq.mp2mr)); ++i) {
>> - if (unlikely(txq_ctrl->txq.mp2mr[i].mp == NULL)) {
>> + struct ibv_mr *mr = txq_ctrl->txq.mp2mr[i].mr;
>> +
>> + if (unlikely(mr == NULL)) {
>> /* Unknown MP, add a new MR for it. */
>> break;
>> }
>> - if (txq_ctrl->txq.mp2mr[i].mp == mp)
>> + if (start >= (uintptr_t)mr->addr &&
>> + end <= (uintptr_t)mr->addr + mr->length)
>> return;
>> }
>> txq_mp2mr_reg(&txq_ctrl->txq, mp, i);
>
> if (start >= (uintptr_t)mr->addr &&
> end <= (uintptr_t)mr->addr + mr->length)
>
> Is this expected to have a memory region bigger than the memory pool
> space? I mean I was expecting to see strict equality in the addresses.
In mlx5_mp2mr(), start/end of a memory region are rounded up to make it
aligned to its hugepage size.
struct ibv_mr *
mlx5_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp)
{
[...]
/* Round start and end to page boundary if found in memory segments. */
for (i = 0; (i < RTE_MAX_MEMSEG) && (ms[i].addr != NULL); ++i) {
uintptr_t addr = (uintptr_t)ms[i].addr;
size_t len = ms[i].len;
unsigned int align = ms[i].hugepage_sz;
if ((start > addr) && (start < addr + len))
start = RTE_ALIGN_FLOOR(start, align);
if ((end > addr) && (end < addr + len))
end = RTE_ALIGN_CEIL(end, align);
}
Thanks,
Yongseok
next prev parent reply other threads:[~2017-07-03 20:54 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-28 23:03 [dpdk-dev] [PATCH 0/5] net/mlx5: add vectorized Rx/Tx burst for x86 Yongseok Koh
2017-06-28 23:03 ` [dpdk-dev] [PATCH 1/5] net/mlx5: change indexing for Tx SW ring Yongseok Koh
2017-06-30 12:20 ` Nélio Laranjeiro
2017-06-28 23:04 ` [dpdk-dev] [PATCH 2/5] net/mlx5: free buffers in bulk on Tx completion Yongseok Koh
2017-06-30 12:30 ` Nélio Laranjeiro
2017-06-30 12:43 ` Nélio Laranjeiro
2017-06-30 17:49 ` Yongseok Koh
2017-06-28 23:04 ` [dpdk-dev] [PATCH 3/5] net/mlx5: use buffer address for LKEY search Yongseok Koh
2017-06-30 13:01 ` Nélio Laranjeiro
2017-06-30 18:58 ` Yongseok Koh
2017-06-28 23:04 ` [dpdk-dev] [PATCH 4/5] net/mlx5: select Rx/Tx callbacks when starting device Yongseok Koh
2017-06-30 13:02 ` Nélio Laranjeiro
2017-06-28 23:04 ` [dpdk-dev] [PATCH 5/5] net/mlx5: add vectorized Rx/Tx burst for SSE4.1 Yongseok Koh
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 0/5] net/mlx5: add vectorized Rx/Tx burst for x86 Yongseok Koh
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 1/5] net/mlx5: change indexing for Tx SW ring Yongseok Koh
2017-07-03 13:48 ` Nélio Laranjeiro
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 2/5] net/mlx5: free buffers in bulk on Tx completion Yongseok Koh
2017-07-03 13:58 ` Nélio Laranjeiro
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 3/5] net/mlx5: use buffer address for LKEY search Yongseok Koh
2017-07-03 14:06 ` Nélio Laranjeiro
2017-07-03 20:54 ` Yongseok Koh [this message]
2017-07-04 6:54 ` Nélio Laranjeiro
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 4/5] net/mlx5: select Rx/Tx callbacks when starting device Yongseok Koh
2017-07-03 13:49 ` Nélio Laranjeiro
2017-06-30 19:23 ` [dpdk-dev] [PATCH v2 5/5] net/mlx5: add vectorized Rx/Tx burst for SSE4.1 Yongseok Koh
2017-07-03 23:54 ` Thomas Monjalon
2017-07-04 8:58 ` Nélio Laranjeiro
2017-07-05 0:38 ` Yongseok Koh
2017-07-05 8:21 ` Nélio Laranjeiro
2017-07-05 17:41 ` Yongseok Koh
2017-06-30 21:28 ` [dpdk-dev] [PATCH v2 0/5] net/mlx5: add vectorized Rx/Tx burst for x86 Bruce Richardson
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 " Yongseok Koh
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 1/5] net/mlx5: change indexing for Tx SW ring Yongseok Koh
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 2/5] net/mlx5: free buffers in bulk on Tx completion Yongseok Koh
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 3/5] net/mlx5: use buffer address for LKEY search Yongseok Koh
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 4/5] net/mlx5: select Rx/Tx callbacks when starting device Yongseok Koh
2017-07-06 7:17 ` Nélio Laranjeiro
2017-07-05 18:12 ` [dpdk-dev] [PATCH v3 5/5] net/mlx5: add vectorized Rx/Tx burst for SSE4.1 Yongseok Koh
2017-07-05 22:58 ` Yongseok Koh
2017-07-06 7:16 ` Nélio Laranjeiro
2017-07-06 9:58 ` Ferruh Yigit
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 0/5] net/mlx5: add vectorized Rx/Tx burst for x86 Yongseok Koh
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 1/5] net/mlx5: change indexing for Tx SW ring Yongseok Koh
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 2/5] net/mlx5: free buffers in bulk on Tx completion Yongseok Koh
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 3/5] net/mlx5: use buffer address for LKEY search Yongseok Koh
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 4/5] net/mlx5: select Rx/Tx callbacks when starting device Yongseok Koh
2017-07-06 18:41 ` [dpdk-dev] [PATCH v4 5/5] net/mlx5: add vectorized Rx/Tx burst for x86 Yongseok Koh
2017-07-07 9:58 ` [dpdk-dev] [PATCH v4 0/5] " Ferruh Yigit
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=4771B0EC-2A1E-4A0C-86C5-231289F3E0F2@mellanox.com \
--to=yskoh@mellanox.com \
--cc=adrien.mazarguil@6wind.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=nelio.laranjeiro@6wind.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).