Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:101958
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain gmail.com designates 209.85.214.54 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <CAEsg9X3iKgtNP7fV9JML2dN5hd_U+rt2wJ_-vT5AjiRoCYEPDw@mail.gmail.com>
References: <CAEsg9X3iKgtNP7fV9JML2dN5hd_U+rt2wJ_-vT5AjiRoCYEPDw@mail.gmail.com>
Date: Thu, 8 Mar 2018 18:52:44 +0100
Message-ID: <CAF+90c-X5Yf_=C-36xgGQ9RrOiJ-eRdjip0qmBv30J92Lv_6BQ@mail.gmail.com>
To: David Rodrigues <david.proweb@gmail.com>
Cc: PHP Internals <internals@lists.php.net>
Content-Type: multipart/alternative; boundary="94eb2c05b8cacd219f0566ea57e2"
Subject: Re: [PHP-DEV] Direct method call slower than method_exists()
From: nikita.ppv@gmail.com (Nikita Popov)

--94eb2c05b8cacd219f0566ea57e2
Content-Type: text/plain; charset="UTF-8"

On Thu, Mar 8, 2018 at 6:41 PM, David Rodrigues <david.proweb@gmail.com>
wrote:

> I have take note that a lot of projects (like Laravel) prefer uses
> method_exists() than implements an empty function (that should be useful to
> IDE when it is not annotated as an @method on class).
>
> For some reason, method_exists() will works faster than direct call when
> method doesn't exists vs. a direct method call when method is empty.
>
> My benchmarks:
>
> Case #1:
> * Direct call, empty method:
> * Cycles by min. : 26.459.804
>
> Case #2:
> * Direct call, simple content (eg. is_bool(true)):
> * Cycles by min. : 24.771.454
> - 6.38% slower than Case #1
>
> Case #3:
> * method_exists(), no method:
> * Cycles by min. : 45.014.690
> - 70.12% faster than Case #1
>
> Case #4:
> * method_exists(), method exists with simple content:
> * Cycles by min. : 18.068.555
> - 27.06% slower than Case #2
>
> You will note that method_exists() will be more useful when method doesn't
> exists (+70% faster), but worse in cases where it does (27% slower). Which
> make that a hard decision, because you need "guess" the method declaration
> usage when implements something like that. If case the method existence is
> very low, then you will prefer method_exists(), which is bad for code
> design.
>
> My suggestion here (that I don't know if could be applied as is): identify
> empty methods and avoiding the execution (once that it will not do anything
> anyway). So we could just implement empty methods without any problem.
>
> Maybe it could be done at this point and related:
> https://github.com/php/php-src/blob/master/Zend/zend_
> language_parser.y#L725
>
> --
> David Rodrigues


Could you please provide the used benchmarking code?

Nikita

--94eb2c05b8cacd219f0566ea57e2--