Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:86481
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: softfail (pb1.pair.com: domain golemon.com does not designate 209.85.215.51 as permitted sender)
MIME-Version: 1.0
Sender: php@golemon.com
In-Reply-To: <556F80CF.40303@moonspot.net>
References: <CAEYWF=4Gn=6OcPWc3w2k6G7DWrqaUfS4QdohjhiSG+2vH8euyg@mail.gmail.com>
	<556F80CF.40303@moonspot.net>
Date: Wed, 3 Jun 2015 16:34:16 -0700
Message-ID: <CAESVnVovWtsfG-rt_A9dehy60QNyjvZAAu9ETXy8H4qeNE_EMg@mail.gmail.com>
To: Brian Moon <brian@moonspot.net>
Cc: Bishop Bettini <bishop@php.net>, PHP internals <internals@lists.php.net>
Content-Type: text/plain; charset=UTF-8
Subject: Re: [PHP-DEV] Method call overhead
From: pollita@php.net (Sara Golemon)

On Wed, Jun 3, 2015 at 3:33 PM, Brian Moon <brian@moonspot.net> wrote:
> On 6/2/15 22:30 , Bishop Bettini wrote:
>> I've measured the overhead for method calls in a variety of environments
>> (Amazon, Travis, and 3v4l).  The results are reliable and here's 3v4l
>> <http://3v4l.org/NsjJR>.
>>
>> Some observations.  First, as expected, direct calls are faster than
>> static
>> object calls, which are faster than object calls.  Second, in absolute
>> times PHP7 outperforms HHVM3 substantially.  Kudos, really impressive.
>>
>> My question though is on relative times.  Method call overhead is
>> consistently 50% to 150% over a direct call.  Is my experiment invalid, or
>> is this overhead expected?  Is the overhead in the allocation,
>> deallocation, GC?
>>
> This is a better representation of what you are trying to show. It removes
> all the magic call back stuff that could be adding to the slowness you are
> seeing. In addition, it does not create a new object on every call for the
> object method. Creating a new object is going to explicitly slow things
> down. But, it's not related to the call time.
>
Interesting data.  Regarding your original question, I would expect
method calls to be somewhat more expensive since even with a known
method name there's polymorphism to take into account.  I wouldn't
expect it to be massively more, but non-zero.

I would still recommend using Callgrind as Mike suggested.  It's going
to give you much more reliable (and useful) time numbers than
microtime().

-Sara