[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <866b75ce-1692-5878-5b98-ec2a12e665bc@huawei.com>
Date: Mon, 4 Nov 2024 15:34:45 +0800
From: Li Qiang <liqiang64@...wei.com>
To: "D. Wythe" <alibuda@...ux.alibaba.com>, <wenjia@...ux.ibm.com>,
<jaka@...ux.ibm.com>, <tonylu@...ux.alibaba.com>, <guwen@...ux.alibaba.com>
CC: <linux-s390@...r.kernel.org>, <netdev@...r.kernel.org>,
<linux-kernel@...r.kernel.org>, <luanjianhai@...wei.com>,
<zhangxuzhou4@...wei.com>, <dengguangxing@...wei.com>, <gaochao24@...wei.com>
Subject: Re: [PATCH] net/smc: Optimize the search method of reused buf_desc
在 2024/11/4 9:41, D. Wythe 写道:
>
>
> On 10/29/24 2:54 PM, liqiang wrote:
>> We create a lock-less link list for the currently
>> idle reusable smc_buf_desc.
>>
>> When the 'used' filed mark to 0, it is added to
>> the lock-less linked list.
>>
>> When a new connection is established, a suitable
>> element is obtained directly, which eliminates the
>> need for traversal and search, and does not require
>> locking resource.
>>
>> A lock-free linked list is a linked list that uses
>> atomic operations to optimize the producer-consumer model.
>
>
>
> No objection, but could you provide us with some data before and after the optimization ?
> .
I have resent this patch a few days ago with '[PATCH net-next]' prefix.
It contains more detailed function time-consuming and nginx test data.
You can find some test data in that email. :)
Let me summarize it here:
1. The function 'smc_buf_get_slot' takes less time when a new SMC link is established,
5us->100ns (when there are 200 active links), 30us->100ns (when there are 1000 active links).
2. Using wrk and nginx to test multi-threaded short connection performance
has significantly improved.
Environment: QEMU emulator version 1.5.3 @ Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz
Test with SMC loopback-ism.
--
Best regards,
Li Qiang
Powered by blists - more mailing lists