Description:
------------
Git repository with code to reproduce bug: https://github.com/LukeMauldin/laravel-perf

The above Git repository is a standard Laravel application with these routes added:
/showview
/simplejson
/simplepdo/{N}
/simpleeloquent/{N}

I obtained the HTTP response times below (in milliseconds) on a AWS Windows Server 2008 R2 instance using IIS and FastCGI.  I measured using curl on the local instance to prevent any network problems.  They were repeatable over a period of several days.  PHP 5.6 is version 5.6.4  PHP 5.4 is version 5.4.4.  Both use the standard PHP Windows downloads from http://windows.php.net/

Url	PHP 5.4 wo/Opache	PHP 5.4 w/Opache	PHP 5.6 wo/Opache	PHP 5.6 w/Opache
/showview	60	17	51	17
/simplejson	60	16	53	16
/simplepdo/1	70	23	59	23
/simplepdo/100	294	130	197	182
/simpleeloquent/1	85	28	67	29
/simpleeloquent/100	365	260	298	272

PHP 5.6 without Opcache is faster than PHP 5.4 without Opcache, which is expected.  However, with Opcache enabled, PHP 5.4 is faster than PHP 5.6 in the /simplepdo/100 and /simpleeqloent/100 tests which is clearly a regression.  The php.ini section for opcache is the same in both:

[opcache]
opcache.enable = 1
opcache.memory_consumption = 64
opcache.interned_strings_buffer = 12
opcache.max_accelerated_files=16629
opcache.revalidate_freq=60
opcache.fast_shutdown=1
opcache.save_comments = 0
opcache.enable_cli=1