Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
In-Reply-To: <6dc928cd-adbb-0304-1308-fb556d48698d@163.com>
References: <20180722173715.25327-1-zengzhaoxiu@163.com> <20180722183706.GA7979@kroah.com>
 <9b18802b-a451-eb81-3317-1de406f03118@163.com> <6dc928cd-adbb-0304-1308-fb556d48698d@163.com>
From:   Andy Shevchenko <andy.shevchenko@gmail.com>
Date:   Fri, 27 Jul 2018 13:39:22 +0300
Message-ID: <CAHp75Vd2Lu110om-hPE2bx_XFa0zpNHHC4RC_uLsTBSywJO2KQ@mail.gmail.com>
Subject: Re: [PATCH 1/1] lib: use sunday algorithm to do strstr() and strnstr()
To:     Zhaoxiu Zeng <zengzhaoxiu@163.com>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Matthew Wilcox <mawilcox@microsoft.com>,
        Dan Carpenter <dan.carpenter@oracle.com>,
        Geert Uytterhoeven <geert@linux-m68k.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Andrey Ryabinin <aryabinin@virtuozzo.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Zhaoxiu Zeng <zhaoxiu.zeng@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Jul 27, 2018 at 8:48 AM, Zhaoxiu Zeng <zengzhaoxiu@163.com> wrote:
> =E5=9C=A8 2018/7/27 1:17, Zhaoxiu Zeng =E5=86=99=E9=81=93:
>> =E5=9C=A8 2018/7/23 2:37, Greg Kroah-Hartman =E5=86=99=E9=81=93:
>>> On Mon, Jul 23, 2018 at 01:37:15AM +0800, Zhaoxiu Zeng wrote:

>>>> The Sunday algorithm is a variation of Boyer-Moore algorithm, it is ea=
sy and fast.
>>>> For the Sunday algorithm, to see
>>>>     http://www.inf.fh-flensburg.de/lang/algorithmen/pattern/sundayen.h=
tm
>>>
>>> So you say, but what does this really buy us?  Why make this change?
>>> How was it tested?  What is the downside of not taking this?

>> I use the following program to test on fc28.
>> Compile with O2, the new version is almost 2X faster than the original.

>> The code size of the original is 0x80, the newer is 0xB0.

So, output of bloat-o-meter would be good to have in commit message.

>> The test result:

Compact performance statistics as well.

>> Thanks!

> The original strnstr might has a bug too!
> For example, assume s1 is "123\0abc...." and s2 is "abc\0",
> call strnstr(s1, s2, 7) will return &s1[4], but the correct result is NUL=
L.

If there is a bug, send another patch to fix the bug first.

--=20
With Best Regards,
Andy Shevchenko