PHP two include load file mode efficiency comparison

Continue to improve the "X plan" core, the core, it is necessary to load the file, tried two methods, found that efficiency is different, share it ~
Let's say two ways:

1 defines a string variable that holds the list of files to be loaded. Then foreach loads.
$a = '/a.class.php;/util/b.class.php;/util/c.class.php ';
$b = '/d.php;/e.class.php;/f.class.php;/g.class.php ';

Load Basic System files
$kernel _require_files = explode ('; ', $a);//sys_require_lib_file_list);
foreach ($kernel _require_files as $f) {
Require_once (Sys_lib_path. ') /system '. $f);
}

Load Basic System files
$kernel _require_files = explode ('; ', $b);//sys_base_file_list);
foreach ($kernel _require_files as $f) {
Require_once (Kernel_path. $f);
}

2 All the files to be loaded in an include file loaded, the current page directly include this include file.
include.php File Contents
Require_once (' func.php ');
Require_once (' LangManager.class.php ');
Require_once (' _kernelautoloader.class.php ');
Require_once (' ApplicationSettingManager.class.php ');

Require_once (' lib/system/activator.class.php ');
Require_once (' lib/system/util/cxml.class.php ');
Require_once (' lib/system/util/cweb.class.php ');

I personally think that the second method is more efficient, because there is no such redundant operation of foreach ~ All things to prove, not a figment of the imagination, so I verified. The following is the time that is consumed by randomly loading 10 times in two ways:
Foreach
0.017754077911377
0.017686128616333
0.017347097396851
0.018272161483765
0.018272161483765
0.018401145935059
0.018187046051025
0.020787000656128
0.018001079559326
0.017963171005249

Include_once (' include.php ');
0.025792121887207
0.024733066558838
0.025041103363037
0.024915933609009
0.024657011032104
0.024134159088135
0.025845050811768
0.024954080581665
0.024757146835327
0.02684497833252

In addition, try again to load all files directly on the current page
0.022285938262939
0.024394035339355
0.023194074630737
0.023229122161865
0.024644136428833
0.023538112640381
0.024240016937256
0.025094032287598
0.023231029510498
0.02339506149292

The result surprised me! Unexpectedly, the first seemingly slowest method, the least time-consuming, and directly on the current page to load more than a few files time-consuming ah ~