Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:82104
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain zeyos.com designates 109.70.220.166 as permitted sender)
Date: Sun, 08 Feb 2015 00:54:20 +0100
To: Oleg Serov <serovov@gmail.com>,
	internals@lists.php.net
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset="UTF-8"
Message-ID: <20150207235422.930125F810@mx.zeyon.net>
Subject: Re: [PHP-DEV] Idea of optimizing php !empty(...) expression.
From: ben.coutu@zeyos.com (Benjamin Coutu)

Hi Oleg,

this is already accomplished.=20

if (! <exp>) <stat>

would usually be compiled to:

1: <exp>
2: NOT
3: JMPZ -> 5
4: <stat>
5: ...

But OpCache's built-in optimizer (https://github.com/php/php-src/blob/maste=
r/ext/opcache/Optimizer/block_pass.c) is able to recognize this during the =
block level analysis and creates the following opcodes for the Zend engine:

1: <exp>
2: JMPNZ -> 4
3: <stat>
4: ...

Please note that NOT+JMPZ are converted to an equivalent single opcode JMPN=
Z. An extra NOT_EMPTY opcode is therefore not necessary.

I do not quite understand why you experience a performance difference, prob=
ably OpCache is not enabled or optimization flags are not properly set. Ple=
ase see http://php.net/manual/en/book.opcache.php for more info.

Cheers,

Benjamin Coutu
=0A=0A=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D Original =3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=0AFrom: Oleg Serov <serovov@gmail.com>=0ATo: internals@lists.php.net=0A=
Date: Sun, 08 Feb 2015 00:13:05 +0100=0ASubject: [PHP-DEV] Idea of optimizi=
ng php !empty(...) expression.=0A=0AI use !empty() very often and decided t=
o make a benchmark test.=0A=0AHere is the code and results: http://pastebin=
.com/fMhhdQiW=0A=0Aif (!empty(...)) working on 23% slower than if (empty())=
 expression.=0A=0ASo if create new operator not_empty() it will improve per=
formance.=0A=0AThe first question is: What do you think about optimizing !e=
mpty(...), do=0Awe need it ?=0A=0AAnd I see two way to make this happen.=0A=
=0A1. Create new language entity "not_empty".=0A2. Improve parser and help =
to handle "!empty" calls different way.=0A=0AIt is obviously that option 2 =
is better. Is it real to optimize parser that=0Away?=0A=0AThanks!