Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:96227
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain cubiclesoft.com designates 149.56.142.28 as permitted sender)
To: Pascal KISSIAN <php-mailing-list@lool.fr>,
 PHP Development <internals@lists.php.net>
References: <004101d21e33$124edf90$36ec9eb0$@lool.fr>
Message-ID: <79b96cf5-dd10-11e9-8b8c-5fd763fb4796@cubiclesoft.com>
Date: Tue, 4 Oct 2016 08:24:02 -0700
User-Agent: Mozilla/5.0 (X11; Linux i686; rv:11.0) Gecko/20120327
 Thunderbird/11.0.1
MIME-Version: 1.0
In-Reply-To: <004101d21e33$124edf90$36ec9eb0$@lool.fr>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] Feature Request: inline pseudo-instruction
From: thruska@cubiclesoft.com (Thomas Hruska)

On 10/4/2016 4:33 AM, Pascal KISSIAN wrote:
> Hi everybody,
>
> I have an application where a small file is included at multiple places.
>
> So far so good.
>
> The problem is that this include consists in a small piece of code which is
> inside a multi-level loop.
>
> The include is done about an average of 100.000 times .
>
> When I manually replace the include with the code which is inside, the
> performance boost is about  a 50-100 factor..
> (with opcache enabled, it is far worst without opcache).

I'm not surprised.

What's wrong with using PHP functions and/or classes?  Use 
'require_once' outside the loops, which brings in the file containing 
the function/class, and then execute the function inside the loop. 
Loading a file repeatedly incurs a lot of processing time to parse it 
into opcodes, which, as you noticed, the opcode cache helps 
significantly with.  You are also already getting the performance 
benefits of the file already being in RAM thanks to whatever OS level 
in-memory file caching system your OS uses.  Basically, your code is 
(ab)using system caching as a performance crutch, which means it is time 
to refactor your code.

Once you switch to using a function, there are additional techniques 
that can be used to speed up the function itself - prefer PHP keywords 
over PHP functions (e.g. array_key_exists() is many times slower than 
'isset'), find an alternate algorithm and/or shortcuts with the current 
algorithm, write a PHP extension, etc.

The 'include' and 'require' keywords invite abuse similar to what you 
just described and, when abused, will have performance metrics similar 
to what you described.  You aren't the first person to use those 
keywords in a loop and likely won't be the last.  The documentation 
should be updated to reflect a preference for the 'require_once' and 
'include_once' keywords, which would help to avoid the misunderstanding 
that no one should use any of the code inclusion keywords in a loop.

Personally, I use 'require_once' exclusively for a wide variety of 
reasons.  I've rarely seen a need to use the other include/require 
keywords.  Most of the time I see 'include', I usually think, "That code 
could be refactored to use 'require_once' and the author would be better 
off for it."

-- 
Thomas Hruska
CubicleSoft President

I've got great, time saving software that you will find useful.

http://cubiclesoft.com/