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;
Reply-To: efremov@linux.com
Subject: Re: [PATCH] scripts: coccinelle: check for !(un)?likely usage
To:     Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        Julia Lawall <julia.lawall@lip6.fr>
Cc:     Joe Perches <joe@perches.com>, cocci@systeme.lip6.fr,
        linux-kernel@vger.kernel.org,
        Gilles Muller <Gilles.Muller@lip6.fr>,
        Nicolas Palix <nicolas.palix@imag.fr>,
        Michal Marek <michal.lkml@markovi.net>
References: <20190825130536.14683-1-efremov@linux.com>
 <b5bae2981e27d133b61d99b08ee60244bf7aabe3.camel@perches.com>
 <88f6e48e-1230-9488-a973-397f4e6dfbb5@linux.com>
 <4E9DDF9E-C883-44F0-A3F4-CD49284DB60D@lip6.fr>
 <95c32d19-eb4d-a214-6332-038610ec3dbd@rasmusvillemoes.dk>
From:   Denis Efremov <efremov@linux.com>
Message-ID: <16053035-655a-7d53-29d1-ea914e3a21dd@linux.com>
Date:   Wed, 28 Aug 2019 15:33:46 +0300
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <95c32d19-eb4d-a214-6332-038610ec3dbd@rasmusvillemoes.dk>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 8/28/19 2:33 PM, Rasmus Villemoes wrote:
> On 25/08/2019 21.19, Julia Lawall wrote:
>>
>>
>>> On 26 Aug 2019, at 02:59, Denis Efremov <efremov@linux.com> wrote:
>>>
>>>
>>>
>>>> On 25.08.2019 19:37, Joe Perches wrote:
>>>>> On Sun, 2019-08-25 at 16:05 +0300, Denis Efremov wrote:
>>>>> This patch adds coccinelle script for detecting !likely and !unlikely
>>>>> usage. It's better to use unlikely instead of !likely and vice versa.
>>>>
>>>> Please explain _why_ is it better in the changelog.
>>>>
>>>
>>> In my naive understanding the negation (!) before the likely/unlikely
>>> could confuse the compiler
>>
>> As a human I am confused. Is !likely(x) equivalent to x or !x?
> 
> #undef likely
> #undef unlikely
> #define likely(x) (x)
> #define unlikely(x) (x)
> 
> should be a semantic no-op. So changing !likely(x) to unlikely(x) is
> completely wrong. If anything, !likely(x) can be transformed to
> unlikely(!x).

As far as I could understand it:

# define likely(x)	__builtin_expect(!!(x), 1)
# define unlikely(x)	__builtin_expect(!!(x), 0)

From GCC doc:
__builtin_expect compares the values. The semantics of the built-in are that it is expected that exp == c.

if (!likely(cond))
if (!__builtin_expect(!!(cond), 1))
if (!((!!(cond)) == 1))
if ((!!(cond)) != 1) and since !! could result in 0 or 1
if ((!!(cond)) == 0)

if (unlikely(cond))
if (__builtin_expect(!!(cond), 0))
if ((!!(cond)) == 0))

Thanks,
Denis